Regulation of Glucose Concentration Using Artificial Pancreas During and after Physical Activity

Transkripcija

1 University of Ljubljana Faculty of Medicine Klemen Dovč, MD Regulation of Glucose Concentration Using Artificial Pancreas During and after Physical Activity Nadzor urejanja ravni glukoze z uporabo umetne trebušne slinavke med telesno aktivnostjo in po njej Ph.D. Thesis Ljubljana, 2017

2 University of Ljubljana Faculty of Medicine Klemen Dovč, MD Regulation of Glucose Concentration Using Artificial Pancreas During and after Physical Activity Nadzor urejanja ravni glukoze z uporabo umetne trebušne slinavke med telesno aktivnostjo in po njej Mentor appointed by the Senate of Faculty of Medicine on: Imenovanje mentorja na seji senata dne: The evaluation and defense commission appointed by the Senate of Faculty of Medicine on: Komisija za oceno in zagovor imenovana na seji senata dne: Date of defense: Datum zagovora: Mentor: Mentor: prof. dr. Tadej Battelino, dr. med. President of the commission: Predsednik komisije: prof. dr. Andrej Janež, dr. med. Member: Član: izr. prof. dr. Janez Jazbec, dr. med. Member: Član: prof. Stuart A. Weinzimer, MD

3 English proofreading: Lektoriranje angleškega besedila: Ana Dovč, dr. med., Eva Dovč, BA Slovenian proofreading: Lektoriranje slovenskega besedila: Jože Faganel Key words: type 1 diabetes, children, adolescents, physical activity, glucose control, and artificial pancreas Ključne besede: sladkorna bolezen tipa 1, otroci, mladostniki, telesna aktivnost, nadzor ravni glukoze, umetna trebušna slinavka

4 Contents 4 Contents 6 List of figures 7 List of tables 8 Abbreviations 9 Extended summary 9 Background 9 Participants and methods 13 Results 13 Cardiorespiratory fitness capacity 13 Glucose control and insulin delivery 15 Lactate and cortisol essay analysis 15 Adverse events 15 Conclusions 17 Answers to research hypotheses 18 Razširjeni povzetek 18 Izhodišča 19 Preiskovanci in metode 22 Statistična analiza 22 Rezultati 22 Kardiorespiracijska telesna pripravljenost 22 Uravnavanje ravni glukoze 24 Analiza vrednosti laktata in kortizola 24 Neželeni dogodki/učinki naprave 24 Zaključki 26 Odgovori na zastavljene hipoteze 27 Introduction 27 Epidemiology 28 Metabolic control of children with type 1 diabetes 29 Type 1 diabetes and physical activity 29 Insulin pump use 30 Continuous glucose monitoring and sensor-augmented insulin pump therapy 30 Suspend at Low and Suspend Before Low functions 30 Artificial pancreas glucose control 37 Artificial pancreas and physical activity 38 Purposes of the thesis 39 The aim of the thesis 39 Methods

5 39 Participants 40 Inclusion criteria 40 Exclusion criteria 41 Procedures 42 Exercise protocol 42 Standardized meals and insulin delivery 43 Devices and Assays 43 Fuzzy logic artificial pancreas DreaMed Glucositter 46 Clinical data of artificial pancreas device DreaMed Glucositter 47 Randomization and masking 48 Statistical analyses 48 Safety monitoring 48 Clinical trial Management and Regulatory Approvals 49 Thesis Results 50 Characteristics of participants 50 Cardiorespiratory fitness capacity 51 Glucose control 51 Time in target range 52 Time in hypoglycemia 54 Time in hyperglycemia 55 Glucose control during exercise period 56 Day two glucose control 57 Glucose control during different exercise protocols 58 Insulin delivery 59 Lactate and cortisol essay analysis 61 Adverse events 62 Discussion 62 Contribution to science 62 Population characteristics and cardiorespiratory fitness capacity 62 Glucose control 63 Challenge related to physical activity 64 Lactate and cortisol analysis 65 Thesis limitations 65 Thesis conclusions 66 Answers to thesis research hypotheses 67 Acknowledgments 68 References

6 List of figures 27 Figure 1A. Median age at type 1 diabetes diagnosis in Slovenia by the observational year. 28 Figure 1B. Median Hba1c levels in Slovenia by the observational year. 28 Figure 2. Multiple adjusted estimates of average Hba1c and probabilities of suboptimal glycemic control. 44 Figure 3. A general description of the MD-Logic artificial pancreas. 44 Figure 4. A sensor-augmented insulin pump components. 46 Figure 5. A shematic description of the control algorithm. 49 Figure 6. Study flowchart. 51 Figure 7. Median sensor glucose during artificial pancreas and SAP insulin delivery. 57 Figure 8A. Median sensor glucose during artificial pancreas and SAP insulin delivery for 55% VO 2max physical activity protocol. 58 Figure 8B. Median sensor glucose during artificial pancreas and SAP insulin delivery for 55/80% VO 2max physical activity protocol. 59 Figure 9. Median blood lactate level at the end of exercise protocol during artificial pancreas and SAP insulin delivery. 59 Figure 10. Median blood lactate level at the end of physical activity for 55% VO 2max and 55/80% VO 2max protocols. 60 Figure 11. Median fasting cortisol level in the morning of the following day during artificial pancreas and SAP insulin delivery. 60 Figure 12. Median fasting blood cortisol in the morning of the following day for 55% VO 2max and 55/80% VO 2max protocols. 6

7 List of tables 32 Table 1. The overview of randomized controlled trials on outpatient use of artificial pancreas in type 1 diabetes for glycemic control by year and length of observation period. 50 Table 2. Participants characteristics. 50 Table 3. Lung function test results and cardiorespiratory fitness capacity. 52 Table 4. Comparison of glucose control during artificial pancreas and SAP insulin delivery for outcome measures of near normoglycemia. 53 Table 5. Comparison of glucose control during artificial pancreas and SAP (control) for the outcome measures of hypoglycemia. 54 Table 6. Comparison of glucose control during artificial pancreas use and SAP (control) insulin delivery for the outcome measures of hyperglycemia. 56 Table 7A. Analysis of blood glucose values during the exercise period and for 2 hours after. 56 Table 7B. Analysis of sensor glucose during the exercise period and four hours after. 57 Table 8. Glucose control on the day following physical activity (7h 3h). 58 Table 9. Comparison of insulin delivery during artificial pancreas and SAP (control) glucose control. 61 Table 10. Hypoglycemic events, rescue carbohydrate requirement and ketonuria events during artificial pancreas and SAP (control) insulin delivery. 7

8 Abbreviations AP API BMI SDS CGM CL CSII DKA ECG HbA1c ISPAD MPC PID SAP SD SMBG VO 2max artificial pancres application programming interface body mass index standard deviation score continuous glucose monitoring closed loop continuous subcutaneous insulin infusion diabetic ketoacidosis electrocardiogram glycated hemoglobin A1c International Society for Pediatric and Adolescent Diabetes model predictive control proportional integral derivative sensor augmented pump standard deviation self-monitoring of blood glucose maximal oxygen consumption 8

9 Extended summary Background Regular physical activity plays a fundamental role in managing type 1 diabetes and has a positive impact on cardiovascular health, insulin requirement, body fitness and general wellbeing. Recent data from large diabetes registries has shown a positive correlation between physical activity and metabolic control. However, physical activity is often associated with an increased risk of glycemic excursions and hypoglycemia, particularly during the activity and the following night. Several strategies have been suggested for limiting this risk, including recommendations for basal insulin adjustments and additional carbohydrate consumption, although these recommendations are inconsistent and based on limited evidence, particularly in the pediatric population. The incorporation of intermittent high-intensity sprints into moderate exercise is associated with less hypoglycemia and is recommended by the International Society for Pediatric and Adolescent Diabetes (ISPAD) as a possible strategy to minimize the risk of hypoglycemia. Artificial pancreas device has recently been used in summer camps and free-living conditions and has proven to be a safe and efficient solution, and is now becoming a part of routine clinical care. The data on artificial pancreas glucose control during physical activity is scarce, particularly in children and adolescents. There have been reports of the frequency of post-exercise nocturnal hypoglycemia and overall hypoglycemia frequency being comparable or lower when using the artificial pancreas compared to Sensor Augmented Pump (device without computer algorithm SAP) as a control. The present thesis evaluated the cardiorespiratory capacity of cohort of children with type 1 diabetes, the safety and efficacy of artificial pancreas insulin delivery during unannounced (to the artificial pancreas algorithm) afternoon physical activity and during the following night, and the exercise-related metabolic response of young people with type 1 diabetes. Participants and methods An open-label (allocation assignment was not masked), randomized, two-arm, crossover, in-hospital clinical trial was conducted at the University Children s Hospital in Ljubljana, Slovenia. The clinical trial was carried out in compliance with the Declaration of Helsinki, Good Clinical Practice, and applicable regulatory requirements. The Slovenian National Medical Ethics Committee and the regulatory authority approved the protocol. All participants and their parents provided written informed assent/consent before the trial initiation. 9

10 The screening visit (visit 1) included informed consent acquisition, detailed physical examination, and the confirmation of inclusion/exclusion criteria. The main inclusion criteria were: age years (inclusive), clinical diagnosis of type 1 diabetes for at least 1 year, at least 3 months of current use of an insulin pump, HbA1c below 9.0% (75 mmol/mol), BMI within normal range for age and sex (± 2 SD) and the absence of other medical conditions (apart from a well-controlled hypothyroidism or coeliac disease). The exclusion criteria encompassed concomitant diseases that could influence the metabolic control or compromise the participants safety, known hypoglycemia unawareness or more than two episodes of severe hypoglycemia with seizures and/or coma within the 6 months prior to the screening, and a history of one or more episodes of diabetic ketoacidosis requiring hospitalization within 1 month prior to the screening. At the baseline visit (visit 2) all participants performed a lung function test, a resting electrocardiogram (ECG) and a cycle ergometer exercise test to determine their maximal oxygen consumption rate (VO 2max ). They were randomly assigned (1:1) to perform a physical activity on two consecutive days using artificial pancreas insulin delivery followed by two days of physical activity using SAP without a computer algorithm (control), or vice versa. Following the run-in period, the order was randomly determined using an automated web-based program with randomly permuted blocks of four. The participants and the investigators analyzing the study data were not masked to treatment during the clinical trial. The participants were trained in the use of the glucose sensor before entering a run-in period. We downloaded the run-in period (at least 5 days) glucose sensor data to derive the initial personalized artificial pancreas system settings and to optimize insulin pump settings, which remained the same for all study visits. The insulin dose during the in-hospital stay was decided by a trained nurse educator in consultation with the participant and, if needed, with the on-call diabetologist. Upon admission, a nurse educator checked the sensor and a backup sensor was inserted if needed. During the in-hospital stay, the sensor calibration was scheduled for three times: at the initiation of the artificial pancreas insulin delivery system (around 15:00 h on day one), before dinner (from 20:00 to 21:00 h) and in the morning of day two (around 7:00 h). If the glucose level was out of range or if there was a discrepancy between the blood and the sensor glucose, the calibration was postponed until the glucose level was stable. During separate in-hospital visits (visits 3 6), the participants performed two different 40 min protocols of afternoon physical activity (the exercise began between 16:30 and 19:30 h) on a cycle ergometer. The physical activity included a moderate intensity (55% VO 2max ) physical activity protocol and a combination of moderate physical activity with incorporated high-intensity (80% VO 2max ) sprints (55/80% VO 2max protocol). In both protocols, participants were instructed to pedal at a steady rate of rev/min for 40 min, at 55% VO 2max load. The load was adjusted in real time based on VO 2max as needed. In the 55/80% VO 2max protocol, high-intensity sprints with a duration of 20 s were incorporated, with intervals of 6 10 min activity at 55% VO 2max between the sprints, for a total of 40 min. 10

11 Throughout the exercise, a continuous ECG was recorded, and the inhaled O 2 and exhaled CO 2 were measured. Capillary blood glucose was checked at the beginning of each exercise session, every 15 min during the exercise and every 30 min for 2 h following the exercise, blood lactate was measured at the beginning and at the end of exercise protocol, and fasting cortisol level was measured in the moring of the following day to determine possible overreaching with the exercise protocol. During hospitalization, all participants received standardized meals containing approximately 1 gram of carbohydrates per kg of body mass for the main meals (lunch, dinner, and breakfast) and about half of this amount for snacks. In both study arms, all meals were covered with manual insulin boluses according to the individual s carbohydrate-to-insulin ratio that remained the same for all study visits. For the SAP insulin delivery, the device was disconnected during the exercise and the basal insulin dose was reduced by 20% for four hours following the exercise session. The»suspend on low«and»suspend before low«functions were turned off. During the artificial pancreas insulin delivery, the use of the pump was uninterrupted; the device was applied from 15:00 h on the day of the exercise to 13:00 h of the following day. The exercise was not announced to the artificial pancreas algorithm. All participants used an identical insulin pump (Paradigm Veo; Medtronic Diabetes, Northridge, CA, USA), a subcutaneous continuous glucose monitoring (CGM) device (Enlite II sensor with MiniLink REAL-Time transmitter; Medtronic Diabetes) and a glucose meter (Contour Link meter; Ascensia HealthCare, Basel, Switzerland). The artificial pancreas algorithm (DreaMed Glucositter; DreaMedDiabetes, Petah Tikva, Israel) used a modified vendor-supplied communication module application programming interface (API) to retrieve glucose/insulin data from the MiniMed Paradigm Veo pump and set the insulin treatment according to a fuzzy-logic-based algorithm. The software version operated on a commercial laptop/tablet computer (ThinkPad T450s; Lenovo, Beijing, China), which had a physical connection to a communication dongle (provided by the manufacturer of the insulin pump). The artificial pancreas requires a patient-specific log file for its operation. This log file includes the treatment settings for an individual that are downloaded (the individual s sensitivity factor, carbohydrates factor and basal insulin settings, based on run-in period data). Once this pre-made log file is loaded by the device (a dedicated laptop in this case), the physician can launch the application, check and approve the settings and insert the pump s serial number for each individual. From then on, the system will automatically connect to the pump and sensor and control them. The exercise protocol was performed on a cycle ergometer (Power Cube LF8.5G with Schiller software; Ganshorn, Niederlauer, Germany). The HbA1c level was determined by the immunochemical method using the Siemens DCA Vantage Analyser and lactate and cortisol with Advia Centaur (Siemens Healthcare, USA). A hypoglycemic event was defined as a blood glucose level below 3.3 mmol/l based on sensor glucose readings, with a minimum duration of 20 min. For this clinical trial, only hypoglycemic 11

12 events meeting the specific criteria listed bellow were reported as an adverse event. All sensor glucose values under 3.3 mmol/l were recorded by the study device and included in the statistical analysis. Severe hypoglycemia was considered a serious adverse event and was defined as glucose under 2.8 mmol/l, accompanied by seizures or loss of consciousness, as per ISPAD guidelines, or if intravenous glucose and/or intramuscular glucagon administration was required. All sensor glucose-detected hypoglycemic events were additionally confirmed with self-monitoring of blood glucose (SMBG). When glucose values fell below 3.3 mmol/l if symptomatic, and when glucose values fell below 2.8 mmol/l, regardless of symptoms, 15 grams rescue carbohydrates were administered as per standard in-hospital procedures and recorded as an adverse event. Hyperglycemia or diabetic ketoacidosis were considered a serious adverse event only if blood glucose rose above 13.9 mmol/l and was associated with low serum bicarbonate (< 15 mmol/l) or low ph (< 7.3) and either ketonaemia (β- hydroxybutyrate level above 3 mmol/l) or ketonuria requiring intravenous treatment. Other hyperglycemic events were not reported as adverse events; however, they were recorded by the study device and included in the final analysis. The objectives of this Ph.D. thesis were to assess the cardiorespiratory fitness capacity of young people with type 1 diabetes, to evaluate efficacy, measured as proportion of time in glucose target range mmol/l ( mg/dl), to evaluate safety, derived from the proportion of time in hypoglycemia below 3.3 mmol/l (60 mg/dl), proportion of time below 3.9 mmol/l (70 mg/dl) and the hypoglycemia event rate of artificial pancreas glucose control when compared to SAP for the time period during an afternoon exercise and the following afternoon/ night based on sensor glucose readings, and to assess the exercise-related metabolic changes in the cohort of young people with type 1 diabetes during an afternoon exercise and in the morning of the following day. The thesis hypotheses were: 1. The cardiorespiratory capacity of cohort of children and adolescents with type 1 diabetes is similar to their healthy peers. 2. The artificial pancreas will increase the proportion of time inside the target range mmol/l ( mg/dl). 3. The artificial pancreas will reduce the proportion of time in hypoglycemia below 3.3 mmol/l (60 mg/dl), proportion of time below 3.9 mmol/l (70 mg/dl) and the hypoglycemia event rate when compared to SAP. 4. Exercise-related metabolic respons will differ when challenged by exercise protocol of moderate physical activity with high-intensity sprints vs. without high-intensity sprints. The statistical analysis was based on the modified intention-to-treat population, defined as all randomly assigned participants with more than 67% sensor measurements. Comparisons between the artificial pancreas and SAP were performed using the paired nonparametric Wilcoxon signed rank test. The power of the nonparametric tests for the primary endpoint was 12

13 based on the results of power simulations (MATLAB 2013b, MathWorks) based on the previous studies. We calculated that the enrolment of 20 participants would provide a power of 90% for detecting a 30% reduction in the proportion of time spent with blood glucose levels below 3.3 mmol/l, at a 0.05 two-sided significance level, assuming a 30% dropout. Results Between 20 th January and 16 th June 2016, 25 eligible participants were recruited through the Slovenian National Diabetes Registry, and 20 (nine female) were included and randomized, performed all the trial-mandated activities, and provided the data for the analysis (the study flow diagram is presented in Figure 6). The baseline characteristics are shown in Table 1 and the data on lung function test and cardiorespiratory fitness capacity in Table 2. The mean age was 14.2 ± 2.0 years, duration of diabetes 8.3 ± 3.2 years, HbA1c 7.7 ± 0.6% (60.0 ± 6.6 mmol/mol), duration of pump therapy 7.4 ± 3.2 years and the total daily insulin dose 0.8 ± 0.2 IU/kg. Cardiorespiratory fitness capacity The participants were of average physical fitness: mean BMI 21.5 ± 4.3 kg/m2; VO 2max 43.3 ± 9.3 ml kg -1 min -1 (36.1 ± 4.0 ml kg -1 min -1 for girls and 49.2 ± 8.1 ml kg -1 min -1 for boys) and the maximal heart rate ± 10.2 beats/min (182.9 ± 11.9 beats/min for girls and ± 7.9 beats/min for boys). Glucose control and insulin delivery The data representing glucose control is shown in Tables 4 9 and Figures 7 8. For the total duration of the study, we obtained 96% of sensor data during the artificial pancreas delivery and 97% SAP (control) delivery. The data from all participants was included in the analysis. The artificial pancreas delivery of insulin increased the median (interquartile range, IQR) proportion of the time spent within the target glucose range of mmol/l when compared to SAP: 84.1% ( ) vs 68.7% ( ), p = This was also true when calculated for the nighttime alone: 92.8% ( ) for artificial pancreas and 73.3% ( ) for SAP, p = The overall median and mean glucose levels were lower during the artificial pancreas glucose control than during SAP (p = and p = , respectively). Similarly, the artificial pancreas glucose control increased the proportion of time spent within the glucose range mmol/l when compared to SAP (26.5% vs 18.3%, p = ). We observed a significant reduction of time in hyperglycemia with artificial pancreas use when compared to SAP. Proportion of time above 10 mmol/l was halved for the whole observational period and the difference was even more pronounced for the night alone when the proportion 13

14 of time above 10 mmol/l was reduced from 22.7% to 7.2%, Consequently High blood glucose index was reduced for the whole observational period and for the night alone. Reduced time spent in hyperglycemia may be important for preventing damage to the developing brain. The proportion of time spent in hypoglycemia below 3.3 mmol/l during the afternoon exercise and the following night, based on sensor glucose readings, was 0.00% for both groups ( % for artificial pancreas and % for SAP, p = ). During the study, six hypoglycemic events were recorded in the artificial pancreas group and 12 in the SAP group (p = ); the participants received rescue carbohydrates on seven occasions in the artificial pancreas group (total of 105 g) and on eight occasions in the SAP group (total of 120 g) (Table 10). There were no significant differences for other variables of hypoglycemia (i.e. proportion of time spent with glucose below 3.9 mmol/l, AUC below 3.3 mmol/l and 3.9 mmol/l, and low blood glucose index). There was no difference in glucose variability between the treatment modalities as measured by mean SD. The improvement in glucose control was achieved with a significantly lower total amount of insulin delivered using the artificial pancreas device (112.6 IU vs IU with SAP, p = ), on account of less basal insulin (80.3 IU via artificial pancreas vs IU via SAP, p = ), with no difference in bolus insulin delivered (36.4 IU via artificial pancreas and 40.3 IU via SAP, p = ) (Table 9). SMBG data on glucose control during exercise and early recovery time is shown in Table 7A and the data on sensor glucose analysis in Table 7B. The active insulin amount at the beginning of the exercise and the blood glucose values at the beginning and end of the exercise were similar for both treatment modalities. The difference between the blood glucose levels at the start and end of exercise was 2.6 ( 4.6 to 1.6) mmol/l for artificial pancreas insulin delivery and 2.2 ( 3.5 to 0.9) mmol/l for SAP (p = ). During the exercise, there was one hypoglycemic event in the artificial pancreas group compared to four in the SAP group. Between 7:00 h and 13:00 h of the day following the physical activity, the proportion of time spent in hypoglycemia was low in both study arms (0.0% for both study groups, p = ). For the proportion of time spent with glucose within the range mmol/l for the second day of observation there was a trends towards favoring artificial pancreas delivery of insulin, but it did not reach statistical significance (72.8% in artificial pancreas vs. 65.5% in SAP, p = ). In the subanalysis that compared the effects of the different physical activity protocols during the time spent at various blood glucose concentrations, we observed a significant improvement in the proportion of time spent within the range mmol/l (p = ), time spent above 13.9 mmol/l (p = ) and median glucose value (p = ) with artificial pancreas glucose control during the 55/80% VO2max protocol, but not during the 55% VO2max protocol (Figure 8A and 8B). The amount of carbohydrates consumed during the study period was 1.14 ± 0.34 g/kg for main meals and 0.55 ± 0.41 g/kg for snacks. 14

15 Lactate and cortisol essay analysis There was no significant difference between the lactate values at the end of physical activity in the two treatment modalities: 1.50 mmol/l ( ) for artificial pancreas and 1.9 mmol/l ( ) for SAP, p = The lactate value was significantly higher at the end of 55/80% VO 2max exercise protocol (2.14 mmol/l ( ), compared to 55% VO 2max (1.23 mmol/l ( ), p < 0.001). There was no difference in fasting cortisol value on the morning following physical activity, regardless of treatment modality or the exercise protocol performed: 362 nmol/l ( ) for artificial pancreas vs. 395 nmol/l ( ) for SAP, p = and 386 nmol/l ( ) for 55/80% VO 2max exercise protocol vs. 371 nmol/l ( ) for 55% VO 2max exercise protocol, p = Adverse events No severe hypoglycemia or other serious adverse events occurred during the study. One participant experienced ketonuria between 1 and 5 mmol/l (Keto-Diabur Test 5000, Roche, Switzerland) one hour before the exercise during the open-loop delivery of insulin during the 55% VO 2max visit protocol. The event was associated with the antecedent set occlusion and hyperglycemia. One participant had a local skin reaction to the CGM adhesive. Another participant broke his wrist playing basketball on the day before his last study visit, but was able to follow the study protocol. Conclusions The cardiorespiratory fitness capacity of cohort of children and adolescents in the present clinical trial was comparable to their healthy peers. Use of artificial pancreas insulin delivery was safe during and after an unannounced (to the artificial pancreas algorithm) physical activity, either moderate or mixed moderate with periodic sprints. The artificial pancreas glucose control substantially increased the time spent within the normal glucose range for 15.4% compared to SAP for the whole observational period. Regardless of the physical activity protocol, the artificial pancreas insulin delivery reduced blood excursions during the night, with 93% of the time spent within the glucose target range of mmol/l for the overnight period. There was no difference for the proportion of time spent in hypoglycemia below 3.3 mmol/l. The reason for this was probably related to the protocol design. Due to meticulous adherence to the in-hospital standard operating procedure for preventing hypoglycemia, implemented by the trained nurse educators during the in-hospital stay, there was consequently an equally short amout of time spent in hypoglycemia in both study groups. 15

16 The overnight hypoglycemia event rate was comparably low for both treatment modalities in the present study, and the amount of time spent in hypoglycemia below 3.3 mmol/l or below 3.9 mmol/l was close to zero. We observed a consistent decline in blood glucose during physical activity, regardless of the treatment modality, with a comparable amount of active insulin at the beginning of the activity. The glucose levels at the beginning of the unannounced physical activity were postprandial (following a snack) and were in most instances within high-end of the normal range. This is in line with the recent recommendations for blood glucose level before a physical activity. In the present clinical trial, the exercise was unannounced to the artificial pancreas system and no precautionary measures were made before the physical activity, as this resembles a real-life situation. For the period of time during and following the exercise in the SAP arm, we incorporated adjustments in the insulin therapy based on ISPAD guidelines. Without a reduction in basal rates after the exercise there would probably be significantly more episodes of hypoglycemia in the SAP arm. The present study tested whether the artificial pancreas device can prevent hypoglycemia during and after an unannounced physical activity. The data showed that there is no need to announce a physical activity to the system. However, for a more prolonged or vigorous physical activity the announcement might be beneficial. The glycemic control of the study population was comparable to the data reported for the population of children with type 1 diabetes in Slovenia and in Europe. The lactate value at the end of 55/80% VO 2max physical activity protocol was increased, when compared to only moderate 55% VO 2max physical activity protocol. This was probably the result of time spent above the aerobic threshold during the high-intensity sprints, which were incorporated into 55/80% VO 2max physical activity protocol, and provides an objective assessment for distinguishing the two physical activity intensities. A similar observation was made in previous studies with adult and pediatric populations. There was no difference in lactate values between the treatment modalities. The fasting cortisol level was unaffected by different physical activity protocols. The intensity for exercise protocol was adjusted in real time based on objective and individual VO 2max in order to prescribe the exercise intensity as individually and accurately as possible. The use of titrated exercise intensity prescription prevents heterogeneous exercise stimuli across participants of different age, gender and cardiorespiratory fitness capacity. We acknowledge that present thesis has some limitations. The clinical trial was conducted in a well-controlled and strictly monitored in-hospital environment. Free-living studies with competition motivation are needed, as are studies where the blood glucose levels are lower at the beginning of the physical activity and without the snack before. There was possible selection bias, as most of the participants that applied for the study were regularly active in their everyday life. We also excluded high-risk individuals with hypoglycemia unawareness or those who recently experienced an episode of severe hypoglycemia, who could potentially have additional benefit in this study, therefore diminishing external validity. 16

17 In conclusion, the data from present thesis demonstrates that the artificial pancreas glucose control was safe and efficacious in increasing the time spent within the desired glucose range, with less insulin delivered and without an increase in hypoglycemia during and after the unannounced (to the artificial pancreas algorithm) physical activity in children and adolescents with type 1 diabetes. Larger studies using the artificial pancreas with physical activity in free-living setting and during competitive sports are warranted, as well as studies incorporating high-risk individuals who would especially benefit from the hypoglycemia risk reduction. Answers to research hypotheses 1. Cardiorespiratory fitness capacity of our cohort of children and adolescents with type 1 diabetes was comparable to their healty peers. 2. The use of artificial pancreas significantly increased the proportion of time spent in the target range mmol/l ( mg/dl) for 15.4% (p = ) compared to SAP during and following physical activity in the population of children and adolescents with type 1 diabetes. 3. The use of artificial pancreas did not significantly lower the risk of hypoglycemia defined as the proportion of time below 3.3 mmol/l (60 mg/dl), proportion of time bellow 3.9 mmol (70 mg/dl) and the number of hypoglycemic events compared to SAP during and following physical activity in the population of children and adolescents with type 1 diabetes. 4. The exercise-related response of lactate level was higher after moderate physical activity protocol with high-intensity sprints compared to the one without, but was not affected by the treatment modality. There was no change in fasting cortisol levels for different exercise protocols and treatment modalities. 17

18 Razširjeni povzetek IzhodišËa Skrb za redno telesno dejavnost je eno temeljnih priporočil za uspešno vodenje sladkorne bolezni tipa 1 (SB1) tako zaradi pozitivnega psihičnega učinka kot tudi zaradi ugodnega vpliva na zdravje srca, ožilja in drugih organskih sistemov. Po najnovejših podatkih mednarodnih registrov za vodenje sladkorne bolezni redna telesna dejavnost ugodno vpliva na presnovno urejenost SB1. Vendar pa je telesna dejavnost pogosto povezana s povečanim tveganjem za hipoglikemijo, predvsem med izvajanjem telesne dejavnosti in v noči po njej. Sodobne smernice za vodenje SB1 priporočajo številne ukrepe za zmanjševanje tveganja, kot so spremembe v odmerkih inzulina in dodaten vnos ogljikovih hidratov. Priporočila pa niso dokončno poenotena in temeljijo na še pomankljivih podatkih, kar predvsem velja za populacijo otrok in mladostnikov. Vključevanje visokoenergijskih šprintov v zmerno telesno dejavnost pa celo lahko zmanjša tveganje za hipoglikemijo in jih Mednarodnjo združenje za diabetes otrok in mladostnikov (angl. International Society for Pediatric and Adolescent Diabetes ISPAD) priporoča kot možno strategijo za zmanjševanje tveganja za hipoglikemijo. Umetna trebušna slinavka (UTS, tudi zaprta zanka) je po podatkih nedavnih kliničnih raziskav, ki so bile izvedene v znotraj in zunajbolnišničnem okolju, varna in učinkovita ter tako postaja del nenadzorovane klinične rabe. Podatki o učinkovitosti UTS pri uravnavanju ravni glukoze pri otrocih in mladostnikih med telesno dejavnostjo so skopi. Po podatkih posameznih kliničnih raziskav pri uporabi UTS ni prišlo do povečanja števila hipoglikemij v noči po telesni dejavnosti, celokupna pogostost hipoglikemij pa je ostala nespremenjena ali pa se je zmanjšala v primerjavi s kontrolno skupino senzor z inzulinsko črpalko brez računalniškega algoritma (angl. Sensor Augmented Pump SAP). Cilji doktorske disertacije so bili: opredeliti kardiorespiracijsko telesno zmogljivost kohorte otrok in mladostnikov s sladkorno boleznijo tipa 1, vrednotiti delovanje sistema UTS pri učinkovitosti, merjeno kot delež trajanja (časa) ravni glukoze v območju med 3,9 in 10 mmol/l, ter varnosti, merjeno kot delež trajanja (časa) v območju hipoglikemije pod 3,3 mmol/l (60 mg/dl), pod 3,9 mmol/l (70 mg/dl) in število hipoglikemičnih dogodkov v času popoldanske vadbe in v dnevu/noči po tej vadbi do 13h naslednjega dne glede na senzorske vrednosti glukoze in končno opredeliti z vadbo povzročeni metabolni odziv v kohorti otrok in mladostnikov s sladkorno boleznijo tipa 1. 18

19 Preiskovanci in metode Klinična raziskava, ki je bila odprta (razporeditev v skupino ni bila zakrita), randomizirana, navzkrižno zasnovana in znotrajbolnišnična, je bila izvedena na Pediatrični kliniki v Ljubljani v skladu s helsinško deklaracijo in s soglasjem Komisije Republike Slovenije za medicinsko etiko in Javne agencije Republike Slovenije za zdravila in medicinske pripomočke. Pridobili smo pisna soglasja vseh preiskovancev in njihovih staršev pred vključitvijo v klinično raziskavo. Ob presejalnem pregledu (obisk 1) so preiskovanci in njihovi starši podpisali privolitev po poučitvi, opravili so temeljit telesni pregled, preverili smo vključitvena in izključitvena merila. Glavna vključitvena merila so bili starost 10 do vključno 17 let, diagnoza SB1 vsaj leto dni pred pričetkom raziskave, pričetek uporabe inzulinske črpalke vsaj 3 mesece pred pričetkom raziskave, vrednost glikoziliranega hemoglobina (HbA 1c ) pod 9 %, indeks telesne mase (ITM) med 5. in 95. percentilo glede na starost in spol. Merila za nevključitev sta bila druga kronična bolezen (razen ustrezno vodenega avtoimunskega vnetja ščitnice in celiakije) ali drugo redno zdravljenje, ki bi lahko pomembno vplivalo na vodenje SB1. Izključitvena merila so bila spremljajoča bolezen ali stanje, ki bi lahko vplivalo na presnovno urejenost, pojavljanje nezavednih hipoglikemij ali več kot dve težji hipoglikemiji s krči ali komo v zadnjih 6 mesecih ali več, diabetična ketoacidoza, ki je zahtevala hospitalizacijo, v mesecu pred presajalnim pregledom. Ob izhodičnem pregledu (obisk 2) so vsi preiskovanci opravili teste pljučne funkcije, elektrokardiogram v mirovanju ter maksimalno obremenitveno testiranje na cikloergometru za določitev njihovega maksimalnega privzema kisika (VO 2max ). Nato smo jih naključno razporedili (v razmerju 1:1) v skupino urejanja ravni glukoze z uporabo UTS in nato z uporabo SAP brez računalniškga algoritma odločanja ali obratno. Vrstni red je bil določen s pomočjo spletnega avtomatiziranega računalniškega programa (randomizirano permutirane skupine po 4). Preiskovanci in raziskovalno osebje niso bili nesezanjeni o načinu zdravljenja med klinično raziskavo. Preiskovanci so bili izurjeni v rabi glukoznega senzora pred vstopov v prehodno (angl. run-in) obdobje. Na podlagi podatkov prehodnega obdobja (vsaj 5 dni) smo postavili izhodiščno nastavitev UTS, prilagojeno posameznemu preiskovancu, in prilagodili nastavitve inzulinske črpalke, ki so ostale nespremenjene za celoten čas trajanja klinične raziskave. Odmerke inzulina v bolnišničnem okolju je prilagajala izkušena sestra edukatorka po dogovoru s preiskovanci in po potrebi s pomočja diabetologa na klic. Pred pričetkom smo preverili glukozni senzor in vstavili dodatnega po potrebi. Med obravnavo smo senzor kalibrirali redno trikrat: ob pričetku delovanja UTS (15h), pred večerjo (med 20h in 21h) in zjutraj ob 7h. Kalibriranje smo odložili ali ponovili, kadar je bil velik razkorak med senzorsko vrednostjo in krvnim sladkorjem. dokler se sladkor ni stabiliziral. Med štirimi ločenimi študijskimi obiski (3 6) so preiskovanci izvedli dva različna protokola popoldanske vadbe (zaćetek vadbe je bil v obdobju med 16:30 in 19:30) v trajanju 40 min na cikloergometru. Prvi protokol vadbe je bil 40 min zmerne telesne dejavnosti (55 % VO 2max ), v drugi protokol (55/80 % VO 2max ) smo zmerni telesni dejavnosti dodali 20 sekund trajajoče visokoenergijske šprinte (80 % VO 2max ). Obremenitev smo med protokolom prilagajali v realnem času glede ne 19

20 merjeno vrednost VO 2max. Preiskovanci so prejeli navodila, da poganjajo kolo s frekvenco obratov na minuto v skupnem času 40 minut in z obremenitvijo 55 % Vo 2max. V testiranje so vključili 2 6 visokoenergijskih (80 % VO 2max ) šprintov. Med šprinti je bilo vsaj 6 minut osnovne obremenitve na 55 % VO 2max namen protokola je bil posnemati vsakdanjo otroško telesno dejavnost z izbruhi intenzivne dejavnosti, ki se prepleta z blago ali zmerno telesno dejavnostjo. Med vadbo smo ves čas beležili EKG, merili vdihani O 2 in izdihani CO 2. Krvni sladkor smo izmerili ob začetku vadbe, po 15 in 30 minutah ter ob koncu vadbe. Po zaključku vadbe so meritev ponavljali na pol ure še 2 uri. Laktat smo izmerili ob začetku in koncu vadbe ter 2 uri po vadbi. Jutranji kortizol smo izmerili zjutraj na dan po vadbi za opredelitev možne preobremenjenosti. Preiskovanci so v času obravnave prejeli standardizirani obrok, prilagojen na telesno težo, ki je vseboval približno 1 gram ogljikovih hidratov na kilogram telesne teže, ob tem je obrok vključeval okvirno 50 % ogljikovih hidratov, 20 % beljakovin ter do 30 % maščob (ne več kot 10 % nasičenih). Ob obroku smo preverili na inzulinski črpalki, da preiskovanci niso imeli več aktivnega inzulina, kritje obroka pa je bilo nespremenjeno. Pred obremenitvijo smo preverili raven glukoze; v primeru vrednosti pod 6 ali nad 14 mmol/l smo obremenitev odložili za vsaj eno uro oz. dokler vrednost v času predvidene obremenitve ni bila v zaželenem območju. V kontrolni skupini SAP smo inzulinsko črpalko med telesno dejavnostjo izklopili, za obdobje 4 ur po vadbi pa smo bazalni inzulin znižali za 20 %. Izklopili smo funkcije za zaustavitev črpalke ob nizkem oz. ob predvidenem nizkem sladkorju. Med uporabo UTS je črpalka ostala vklopljena. UTS je bila vklopljena od 15h do 13h naslednjega dne. Telesne dejavnosti računalniškemu algoritmu UTS nismo napovedali. V času klinične raziskave so vsi preiskovanci uporabljali enake inzulinske črpalke (Paradigm Veo; Medtronic Diabetes, Northridge, CA, ZDA), podkožni CGM (Enlite II senzor with MiniLink REAL -Time transmitter; Medtronic Diabetes, ZDA) in merilce krvnega sladkorja (Contour Link meter; Ascensia HealthCare, Basel, Švica). Sistem UTS DreaMed Glucositter je računalniški program (verzija ), ki sprejema podatke iz rutinskega komercialnega senzorja za neprekinjeno podkožno merjenje glukoze in se na podlagi informacij o posameznem bolniku, zbranimi iz spomina senzorja za merjenje glukoze, in na podlagi sprotnih podatkov o učinkovanju prejšnjih odmerkov inzulina odloči za trenutni odmerek inzulina in ga sporoči komercialni inzulinski črpalki. V primeru, da glede na nastavljene parametre algoritem predvidi hipoglikemijo, ustavi dovajanje inzulina in sproži alarm. Tudi sicer je delovanje varovano s številnimi alarmi, ki opozarjajo na možne zaplete. Programska oprema je bila zasnovana s pomočjo Matlab (MathWorks Co., Natick, MA, USA) in vsebuje naslednje komponente: a) komunikacijski modul, b) uporabniški vmestnik, c) kontrolni algoritem in č) opozorilni modul (Slika 3). 20

21 UTS podpira povezavo z napravo Paradigm Veo (Medtronic Co.). Ta vključuje inzulinsko črpalko, senzor za neprekinjeno merjenje glukoze v medceličnini (podkožni senzor, imenovan SOF senzor ali ENLITE senzor), Minilink oddajnik, ComLink (brezžična, nizkoenergijska, radiofrekvenčna povezava, primerna za EU in ZDA frekvence) in RS232 USB pretvorni kabel (Slika 4). Senzor neprestano meri vrednost glukoze v medceličnini, ki jih posreduje Minilinku preko brezžične, nizkoenergijske, radiofrekvenčne povezave sistemu Paradigm Veo vsakih 5 minut, skupaj do 6 dni. Senzor se umerja (kalibrira) z izmerjenimi vrednostmi krvnega sladkorja (Contour Link Bayer). Minilink se lahko uporabi za več zaporednih obdobij oz. menjav senzorja. Paradigm Veo prikazuje in shranjuje vrednosti glukoze, ki jih posreduje Minilink. Inzulinska črpalka je povezana z računalnikom preko Comlinka (brezžična, nizkoenergijska, radiofrekvenčna povezava). UTS uporablja samostojno razvit programski vmesnik, ki pridobiva podatke za vrednosti glukoze in prilagaja odmerjene vrednosti inzulina sistemu Paradigm Veo glede na odločitve računalniškega algoritma. Celotno delovanje UTS je med klinično raziskavo nadzoroval zdravnik. Hipoglikemični dogodek se je definiral kot senzorska vrednost glukoze pod 3,3 mmol (60 mg/dl) z minimalnim trajanjem vsaj 20 min. V izvedeni klinični raziskavi smo kot neželeni dogodek poročali le hipoglikemične dogodke, ki so izpolnjevali specifična merila, navedena spodaj. Vse senzorske vrednosti pod 3,3 mmol/l smo zabeležili in jih vključili v statistično analizo. Huda hipoglikemija se je obravnavala kot resen neželen dogodek. V skladu z mednarodnimi smernicami združenja ISPAD se opredeli kot vrednost krvnega sladkorja pod 2,8 mmol/l s pridruženimi krči in/ali izgubo zavesti, ki je zahtevala intravensko zdravljenje z glukozo in/ali vbrizganje glukagona v mišico. Vse s senzorjem zabeležene hipoglikemije smo potrdili z meritvijo krvnega sladkorja. Kadarje bila vrednost krvnega sladkorja pod 3,3 mmol/l s pridruženimi simptomi ali kadar je bila vrednost krvnega sladkorja pod 2,8 mmol/l ne glede na simptome, so preiskovanci prejeli 15 gramov rešilnih ogljikovih hidratov v skladu s stardandnimi znotrajbolnišničnimi ukrepi, kar smo zabeležili kot neželeni dogodek. Hiperglikemija ali diabetična ketoacidoza se je obravnavala kot resen neželen dogodek kadar je bil sladkor nad 13,9 mmol/l in je bil ob tem znižan serumski bikarbonat (pod 15 mmol/l) ali ph (pod 7,3) s ketonemijo (vrednost β- hydroxybutyrata nad 3 mmol/l) ali ketonurijo, ki je zahtevala intravensko zdravljenje. V doktorski disertaciji smo preverjali naslednje hipoteze: 1. Kardiorespiracijska telesna zmogljivost kohorte otrok in mladostnikov s sladkorno boleznijo tipa 1 je primerljiva z zdravo populacijo. 2. UTS bo povečala delež časa v območju med 3,9 in 10 mmol/l v primerjavi s SAP 3. UTS bo zmanjšala delež časa v območu hipoglikemije pod 3,3 mmol/l (60 mg/dl), pod 3,9 mmol/l (70 mg/dl) in število hipoglikemičnih dogodkov v primerjavi s SAP. 4. Z vadbo povzročeni metabolni odziv se bo razlikoval med obema različnima protokoloma vadbe (zmerna telesna dejavnost z visoenergijskimi šprinti ali brez njih). 21

22 Statistična analiza Pri analizi podatkov smo, ker je bila predvidena navzkrižna zamenjava med skupinima, uporabili parni neparametrični Wilcoxon Sign Rank test. Za izračun moči primarnega izhodišča (čas v območju glukoze v medceličnini pod 3,3 mmol/l) smo izvedli predhodno simulacijo (MATLAB 2011b, Mathworks). Uporabili smo podatke odčitkov senzorjev 102 preiskovancev na standardnem zdravljenju doma. Naša hipoteza je bila, da bomo uspeli z uporabo UTS zmanjšati čas, preživet v območju hipoglikemije pod 3,3 mmol/l v povprečju za 30 % časa. Za zaznavo take razlike z 90-odstotno statistično močjo (α = 0,05) in ob predvidenem 30-odstotnem osipu preiskovancev smo morali vključiti vsaj 20 preiskovancev. Rezultati Med 20. januarjem in 16. junijem 2016 smo preko Slovenskega nacionalnega registra sladkorne bolezni k sodelovanju povabili 25 preiskovancev, v klinično raziskavo smo jih vključili 20 (devet deklet). Vsi vključeni preiskovanci, ki so bili randomizirani, so dokončali vse zahtevane študijske obveznosti in so prispevali podatke za analizo (Slika 6). Osnovne lastnosti preiskovane populacije so predstavljene v Tabeli 2. Povprečna starost preiskovancev je bila 14,2 ± 2,0 let, trajanje SB1 8,3 ± 3,2 let, povprečni HbA1c 7,7 ± 0,6 % (60,0 ± 6,6 mmol/mol), trajanje zdravljenja z inzulinsko črpalko 7,4 ± 3,2 let in inzulinsko razmerje 0,8 ± 0,2 U/kg. Povprečni BMI (21,5 ± 4,3 kg/m2) je bil znotraj normale. Kardiorespiracijska telesna pripravljenost Ob drugem obisku so preiskovanci izvedli teste pljučne funkcije ter maksimalni obremenitveni test (Tabela 3). Njihova kardiorespiracijska telesna pripravljenost je bila povprečna glede na splošno populacijo: VO 2max 43,3 ± 9,3 ml/kg -1 min -1 (36,1 ± 4,0 ml/kg -1 min -1 za dekleta ter 49,2 ± 8,1 ml/kg -1 min -1 za fante), maksimalni srčni utrip je bil 186,6 ± 10,2 udarcev/minuto (182,9 ± 11,9 /min za dekleta 189,6 ± 7,9 /min za fante). Ob zaključku maksimalne obremenitve so izmerili vrednost laktata 7,8 ± 2,2 mmol/l (8,8 ± 2,4 mmol/l za fante in 6,9 ± 1,7 mmol/l za dekleta). Uravnavanje ravni glukoze Podatki o uravnavanju ravni glukoze so predstavljeni v Tabelah 4 9 in Slikah 7 8. Za celotno obdobje klinične raziskave smo pridobili 96 % podatkov senzorja za čas pod nadzorom UTS ter 97 % za čas pod nadzorom SAP. UTS je pomembno povečala delež časa v zaželenem območju vrednosti glukoze med 70 in 180 mg/dl (3,9 10 mmol/l) za celotno obdobje klinične raziskave (mediana in intrakvartilni interval): 84,1 % (70,0 85,5) za UTS in 68,7 % (56,0 77,7) za SAP, p = 0,006, kot tudi samo za nočni čas: 92,8 % (69,8 98,4) za UTS in 73,3 % (61,3 84,2) za SAP, p = 0,008. Mediana koncentracije glu- 22

23 koze za celotno obdobje klinične raziskave je bila nižja pod nadzorom UTS (p = 0,009). Podobno je uporaba UTS povečala delež časa v območju 4,4 6,7 mmol/l v primerjavi s SAP (26,5 % proti 18,3 %, p = 0,0111). Delež časa v območju hipoglikemije pod 60 mg/dl (3,3 mmol/l) po nenapovedani popoldanski vadbi in v noči po njej za celotno obdobje klinične raziskave je bil 0,00 % za obe študijski skupini (0,00 0,76 % za UTS in 0,00 1,06 % za SAP, p = 0,791). V času celotne študije je bilo pod nadzorom UTS zabeleženih šest hipoglikemičnih dogodkov in 12 pod nadzorom SAP, v 7 primerih so prejeli rešilne OH pod nadzorom UTS (skupaj 105 g) ter v 8 pod nadzorom SAP (skupaj 120 g) (Tabela 10). Pod nadzorom UTS so preiskovanci preživeli krajši čas v območju nad 180 ml/dl (10,0 mmol/l, p = 0,023) in nad 250 mg/dl (13,9 mmol/l), p = 0,010. Indeks visokih vrednosti glukoze (HBGI) je bil nižji pod nadzorom UTS (3,6 v primerjavi z 6,2 pod nadzorom SAP, p = 0,008). V ostalih parametrih hipoglikemije (delež časa pod 3,9 mmol/l, p = 0.381), AUC pod 3,3 mmol/l (p = 0,339), AUC pod 3,9 mmol/l (p = 0,687) in v indeksu nizkih vrednosti glukoze (LBGI) (p = 0,658) ni bilo statistično pomembnih razlik. Do izboljšanja glikemične urejenosti je prišlo z manj dovedenega inzulina z uporabo UTS v primerjavi s SAP (112,6 IU za UTS in 203,7 IU za SAP, p = 0,0123), predvsem na račun manj dovedenega bazalnega inzulina (80,3 IU za UTS in 155,2 IU za SAP, p = 0,0074). Razlik v bolusnem inzulinu ni bilo (36,4 IU za UTS in 40,3 IU za SAP, p = 0,8228) (Tabela 9). Podatki meritev krvnega sladkorja med telesno dejavnostjo in v krajšem obdobju okrevanja po njej so prikazani v Tabeli 7A, senzorski podatki pa v Tabeli 7B. Količina aktivnega inzulina vrednost krvnega sladkorja ob začetku vadbe sta bila primerljiva med opazovanima skupina. Sprememba v vrednosti krvnega sladkorja med začetkom in koncem vadbe je bila enaka za obe opazovani skupini: 2,6 ( 4,6 to 1,6) mmol/l za UTS in 2,2 ( 3,5 to 0,9) mmol/l za SAP (p = 0,1000). Med protokolom telesne dejavnosti je prišlo do enega hipoglikemičnega dogodka pod nadzorom UTS in do štirih pod nadzorom SAP. Za obdobje od 7h do 13h na dan, ki je sledil telesni dejavnosti, razlika v deležu časa v območju 3,9 10 mmol/l ni bila statistično značilna (72,8 % za UTS in 65,5 % za SAP, p = 0,0569). Delež časa v območju hipoglikemije je bil 0,0 % za obe opazovani skupini (p = 0,1094). V dodatni analizi podatkov za posamezni protokol smo opažali pomembno povečanje časa znotraj tarčnega območja 3,9 10 mmol/l (p = 0,021), časa nad 13,9 mmol/l (p = 0,023) in mediane glukozne vrednosti (p = 0,015) z uporabo UTS za protokol 55/80 % VO 2max, ne pa za protokol telesne dejavnosti 55 % VO 2max (Slika 8A in 8B). Količina zaužitih ogljikovih hidratov za celotno opazovano populacijo je bila 1,14 ± 0,34 g/kg za glavne obroke in 0,55 ± 0,41 g/kg za prigrizke. 23

24 Analiza vrednosti laktata in kortizola Vrednost laktata ob zaključku protokola 55/80 % VO 2max je bila pomembno povišana v primerjavi s protokolom 55 % VO 2max : 2,14 mmol/l (1,56 2,70) proti 1,23 mmol/l (1, ), p < 0,001. Med različnima skupina dovajanja inzulina ni bilo sprememb v vrednosti laktata na koncu telesne dejavnosti: 1,50 mmol/l (1,19 2,13) za UTS in 1,9 mmol/l (1,12 2,66) za SAP, p = 0,109. Vrednosti jutranjega kortizola se niso razlikovale glede na način dovajanja inzulina (362 nmol/l ( ) za UTS in 395 nmol/l ( ) za SAP, p = 0,290) ali glede na tip vadbe (386 nmol/l ( ) za protokol 55/80% VO 2max in 371 nmol/l ( ) za protokol 55% VO 2max, p = 0,785. Neželeni dogodki/učinki naprave V času klinične raziskave ni prišlo do hude hipoglikemije ali drugega resnega neželenega dogodka (Tabela 10). Pri eni udeleženki smo zabeležili prisotnost ketonov v urinu med 1 in 5 mmol/l (Keto- Diabur Test 5000, Roche, Switzerland) eno uro pred vadbo v SAP skupini pred protokolom 55 % VO 2max, za katere smo predvidevali, da so posledica nedelovanja sistema za dovajanje inzulina in zato hiperglikemije. Obdobje ketoze smo izključili iz končne analize. Pri enemu udeležencu smo opažali lokalno kožno reakcijo na adheziv po namestitvi glukoznega senzorja. En udeleženec si je zlomil zapestje na desni roki med igranjem košarke v domačem okolju en dan pred zadnjim obiskom in je zaključil klinično raziskavo po protokolu. ZakljuËki Presnovna urejenost kohorte otrok in mladostnikov, vključenih v klinično raziskavo, je bila primerljiva s splošno populacijo v Sloveniji in drugod po Evropi. Kardiorespiracijska zmogljivost kohorte otrok in mladostnikov s sladkorno boleznijo tipa 1, vključenih v klinično raziskavo, je bila primerljiva z njihovimi zdravimi sovrstniki. V celotnem opazovanem obdobju smo opažali pomembno povečanje deleža časa znotraj zaželenih vrednosti glukoze med 3,9 in 10 mmol/l za 15,4 % z uporabo UTS v primerjavi s SAP, v nočnem obdobju je bila raven glukoze 93 % časa v zaželenem območju. Za uravnavanje glukoze je bilo celokupno dovedenega manj inzulina v času uporabe UTS v primerjavi z SAP, predvsem zaradi znižane potrebe po bazalnem inzulinu. Čas v območju hiperglikemije nad 10 mmol/l se je z uporabo UTS zmanjšal. Delež časa nad 10 mmol/l za celotno opazovano obdobje se je razpolovil. Razlika je bila še bolj opazna za nočno obdobje, za katero smo opažali zmanjšanje s 22,7 % na 7,2 %. Zmanjševanje časa v hiperglikemiji je lahko pomembno za preprečevanje tveganja za poškodbo razvijajočih se otroških možganov. Med opazovanima skupina ni bilo razlike v deležu časa v hipoglikemiji. Razlog je bil najverjetneje v napačni zasnovi klinične raziskave. Sistematična implementacija standardiziranih znotraj- 24

25 bolniščnih ukrepov za preprečevanje hipoglikemije, ki so jih izvajale oddelčne sestre in sestre edukatorke, je zmanjšalo delež časa, ko je bil preiskovanec v območju hipoglikemije, ki je bil tako majhen v obeh opazovanih skupinah. Število hipoglikemičnih dogodkov v nočnem času je bilo nizko v obeh skupinah in delež časa v območju hipoglikemije pod 3,3 in 3,9 mmol/l je bil blizu vrednosti nič. Beležili smo stalen padec ravni glukoze v času telesne dejavnosti ne glede na način dovajanja inzulina. Vrednost aktivnega inzulina ob začetku vadbe je bila primerljiva med opazovanima skupina, vrednosti glukoze so bile postprandialne (po prigrizku) in v zgornjem območju zaželenih, kar je skladno z mednarodnimi priporočili. V izvedeni klinični raziskavi računalniški algoritem v UTS ni bil opozorjen na telesno dejavnost, prav tako nismo izvajali drugih dodatnih ukrepov pred telesno dejavnostjo, saj to posnema stvarno življensko dogajanje te populacije otrok in mladostnikov. Skladno s priporočili mednarodnega združenja ISPAD smo zmanjšali dovajanje bazalnega inzulina med vadbo in za obdobje 4 ur po njej. Brez tega ukrepa bi najverjetneje beležili večje število hipoglikemičnih dogodkov po vadbi v kontrolni skupini SAP. V klinični raziskavi smo preverjali, ali je UTS uspešna pri preprečevanju hipoglikemije med vadbo in po njej brez napovedi telesne dejavnosti. Podatki kažejo, da pri izvedenemu tipu vadbe napoved ni potrebna. Ob daljši in bolj intenzivni obremenitvi pa bi bilo opozorilo o načinu vadbe lahko koristno. Vrednost laktata je bila povišana po mešanem protokolu vadbe 55/80 % VO 2max v primerjavi z zmerno telesno dejavnostjo 55 % VO 2max, kar je najverjetnega posledica dejstva, da so bili preiskovanci v času visoko-energijskega šprinta nad aerobnim pragom, s čimer smo lahko objektivno razlikovali med različnima intenzitetama vadbe. Glede na način dovajanja inzulina ni bilo spremembe vrednosti laktata. Vrednost jutranjega kortizola je bila nespremenjena ne glede na tip vadbe ali način dovajanja inzulina. Obremenitev med telesno dejavnostjo smo prilagajali v realnem času glede na objektivno in posamezniku lastno vrednost VO 2max, da bi bila obremenitev za posameznega preiskovanca čimbolj natančna. S tem smo preprečili raznolikost telesnih odzivov, ki jo ima telesna vadba lahko na posameznike različne starosti, spola in telesne zmogljivosti. Zavedamo se, da ima doktorska disertacija določene omejitve. Klinična raziskava je bila zasnovana in izpeljana v varnem znotrajbolnišničnem okolju. Potrebne so dodatne klinične raziskave v domačem okolju in s telesno dejavvnostjo, pri kateri bo prisoten tudi tekmovalni naboj, a tudi telesna dejavnost z nižjo začetno vrednostjo krvnega sladkorja in brez dodatnega prigrizka. Možna je bila pristranskost pri izbiri preiskovancev, saj so bili preiskovanci, ki so se prijavili za klinično raziskavo, tudi sicer redno telesno dejavni v domačem okolju. Izključili smo posameznike s povečanim tveganjem za hipoglikemijo ali z nezavednimi hipoglikemijami, ki bi jih izsledki klinične raziskave lahko dodatno uporabilii. 25

26 Podatki doktorske disertacije kažejo, da je uravnavanje ravni glukoze z uporabo UTS varno in učinkovito pri povečanju deleža časa znotraj zaželenih vrednosti, z manj dovedenega inzulina in brez povečanja tveganja za hipoglikemijo med nenapovedano (računalniškemu algoritmu) telesno dejavnostjo in po njej pri otrocih in mladostnikih s sladkorno boleznijo tipa 1. Potrebne so obsežnejše klinične raziskave z uporavo UTS med telesno dejavnostjo zunaj bolnišnic in v tekmovalnem okolju, kot tudi klinične raziskave, ki bodo vključevale posameznike z večjih tveganjem za hipoglikemijo, ki bi jim lahko zmanjšanje izpostavljenosti za hipoglikemijo dodatno koristilo. Odgovori na zastavljene hipoteze 1. Kardiorespiracijska telesna zmogljivost kohorte otrok in mladostnikov s sladkorno boleznijo tipa 1 je bila primerljiva z zdravo populacijo. 2. Uporaba UTS je pomembno povečala delež časa v območju med 3,9 in 10,0 mmol/l ( mg/dl) za 15,4 % (p = 0,0057) v primerjavi s SAP za celotno opazovano obdobje. 3. Uporaba UTS ni zmanjšala delež časa, ko je bil preiskovanec v območju hipoglikemije pod 3,3 mmol/l (60 mg/dl), pod 3,9 mmol/l (70 mg/dl), in števila hipoglikemičnih dogodkov v primerjavi s SAP. 4. Z vadbo povzročeni odziv laktata je bil višji po protokolu zmerne telesne vadbe z vključenimi visokoenergijskimi šprinti v primerjavi s protokolom brez njih. Glede na način dovajanja inzulina ni bilo razlik. V vrednosti jutranjega kortizola ni bilo sprememb, ne glede na način dovajanja inzulina ali ne glede na tip vadbe. 26

27 Introduction Epidemiology Type 1 diabetes is one of the most common chronic diseases that occur in children, and it affects over half a million children below the age of 15 years worldwide. Both the prevalence and the incidence of type 1 diabetes have globally increased significantly based on the data from several international registries, with an annual increase in incidence ranging from 2.72% (SEARCH, United States) (1) and 2.8% (EURODIAB (Europe) (2) to 3.5% DIAMOND (worldwide) (3). Recent data from one of the largest registries showed the incidence reaching an average of 21.7 new cases per youths (4,5). A similar trend was observed in the population of Slovenian children. Under the age of 15 years, the incidence was estimated to be 14.6/ with about newly diagnosed children each year (6). The highest increase was observed in the youngest age group (0 4 years), which is comparable with most Eastern and Central European countries (2) and differs to that reported in the USA (5). The average age at the onset of type 1 diabetes in Slovenia has decreased from 11.9 to 8.9 years in the past decade (Figure 1A) (7). The reported incidence of type 1 diabetes differs greatly between the different geographical regions, from 0.1 per /year in South America (3) to 64.2 per /year in Finland (8). At the same time the life expectancy of people with type 1 diabetes has increased with reduced all-cause mortality, mostly due to the implementation of intensified insulin therapy and the improved metabolic control (9). If present trends continue, a doubling of new cases of type 1 diabetes in children younger than 5 years is predicted between 2005 and 2020, and a prevalence of children younger than 15 years with type 1 diabetes could rise by as much as 70% (2). Figure 1A. Median age at type 1 diabetes diagnosis in Slovenia by the observational year. 27

28 Metabolic control of children with type 1 diabetes An improvement in glycemic control in people with type 1 diabetes substantially reduces their risk of microvascular complications and cardiovascular disease (9 11). In our center, we observed a steady decline of Hba1c for the whole population of 886 children and adolescents from 9.3% (78 mmol/mol) in year 2000 to 7.6% (61 mmol/mol) in year 2016 (Figure 1B). Figure 1B. Median Hba1c levels in Slovenia by the observational year. The predictive model based on a multivariable regression analysis showed that HbA1c level initially decreased as the age increased. However, the relationship was inverted during adolescence with a trend of higher values of HbA1c towards the end of adolescence. Afterwards, HbA1c levels again started to decrease with age. HbA1c level increased with the duration of type 1 diabetes until about the sixth year, after which the glycemic control remained stable. Additionally, both high and low body mass index (BMI) SD score (SDS) were more likely to have suboptimal glycemic control (Figure 2). Figure 2. Multiple adjusted estimates of average Hba1c and probabilities of suboptimal glycemic control (HbA1c above 7.5%) by body mass index (BMI) SD score (SDS). The values on the y-axes can be interpreted as estimated differences in HbA1c from the starting point. 28

29 Type 1 diabetes and physical activity Regular physical activity is, due to its positive impact on cardiovascular health through endothelial function and lipid profiles, insulin requirements, body fitness, and general well-being (20 22) a foundational part of type 1 diabetes management (23). The physical fitness capacity of children with diabetes is reported to be equall to their healthy peers (24), although some data shows that cardiorespiratory fitness of children with type 1 diabetes is lower when compared with healthy controls (25,26). Recent data from a cross-sectional study based on Swedish pediatric diabetes quality registry of 4,655 children and adolescents has also shown a positive influence of physical activity on metabolic control. No matter the gender and the age there was a decline in average Hba1c level associated with increased physical activity (27). However, physical activity is often associated with increased risk of glycemic excursions and hypoglycemia, particularly during exercise and the following night (29 31). Several strategies have been suggested with the aim to limit this risk, including suggestions on basal insulin adjustments and additional carbohydrates consumption, but are inconsistent and based on limited evidence, especially in the pediatric population (32 37). Consequently, patients with type 1 diabetes are less active than recommended: less than 20% of patients manage to do aerobic exercise more than two times per week, and about 60% of patients do not engage in structured exercise at all (38). Insulin pump use The administration of insulin with an insulin pump or continuous subcutaneous insulin infusion (CSII) was introduced almost 40 years ago (39). The insulin pump infuses a rapid-acting human insulin analog to the subcutaneous tissue at slow and steady basal rates to match the individual s needs, with additional bolus doses to cover meals and correct hyperglycemia (40). In the decades of insulin pump use, it has been shown that its use is associated with better glycemic control and lower rates of severe hypoglycemia and diabetic ketoacidosis (DKA) (41,42). Additionally, a recent study showed reduced cardiovascular mortality (43). The insulin pump is considered a safe and effective treatment option for all pediatric age groups and with the availability of reimbursements is its use steadily increasing (7,44,45). This form of therapy has become the insulin regimen of choice in many countries, particularly for the very young (46 48). 29

30 Continuous glucose monitoring and sensoraugmented insulin pump therapy The day-to-day glycemic control is typically delivered via SMBG in an effort to provide patients with type 1 diabetes a reliable guidance for decisions on insulin dosing to prevent and correct glucose excursions. However, SMBG pinpoints the blood glucose level at a particular moment in time and thus fails to capture ongoing glycemic fluctuations that can potentially be associated with damaging the developing brain (49,50). The lack of consistent SMBG monitoring is aggravated by its sheer day-to-day burden, especially in children and adolescents. The modern real-time CGM provides patients with a series of interstitial glucose measurements at intervals of 1 5 min that can be used for real-time adjustments of the treatment regimen. CGM can help patients, their families and caregivers as well as clinicians to make better-informed decisions on how to control blood glucose levels, but only when it is fully adopted in the day-to-day care (51,52). The improvements in recent years have allowed for better accuracy and simplicity of CGM use and, consequently, more successful implementation (53), effective also with non-adjuvant use (54,55). It can be offered to children and adolescents, if the insurance reimbursement is available, given its prohibitive cost (56). Suspend at Low and Suspend Before Low functions Real-time CGM can be used to interrupt insulin delivery at a certain low sensor glucose value (threshold), rather than just sounding an alarm (which can be overheard) and relying on the patient or a family member to take the necessary action. This simple and straightforward algorithm was shown to be very effective in reducing the risk of nocturnal hypoglycemia (57,58). Rather than waiting for the observed glucose to fall below a specified threshold glucose level, one can predict when such an event is likely to happen within a short period. There are several algorithms for making this prediction. In a recent clinical trial that included 100 children and adolescents with type 1 diabetes, the use of a predictive algorithm reduced the number of hypoglycemic events (59). The use of SAPs with a predictive low-glucose insulin-suspend function could reduce the risk of hypoglycemia after physical activity (60). However, there is a limited effect on reducing time in hyperglycemia (59). Artificial pancreas glucose control Over the last years, technological development revealed new possibilities in type 1 diabetes management. The artificial pancreas (AP) device (also closed loop, CL) is the current state-ofthe-art technology, combining CGM, which provides real-time interstitial glucose values, with a modern insulin pump and sophisticated computer algorithm, which directs insulin (single-hormone) or insulin and glucagon (dual-hormone or bihormonal system) delivery in response to the sensor glucose data. The autonomous, graduated modulation of insulin (and glucagon) delivery 30

31 below and above the amount preset in the insulin pump in a glucose-responsive manner differentiates artificial pancreas from conventional insulin pump therapy with or without the function of low-glucose suspend or suspend before low (61). There is much discussion about the pros and cons of the two main hormonal approaches being developed. The inclusion of glucagon in dual-hormone systems more closely mimics the way in which blood glucose levels are controlled naturally, as the actions of these two hormones are complementary. Although dual-hormone glucose control may offer a way of achieving tighter glycemic control and avoiding more hypoglycemia, this comes at the expense of increased cost and device complexity, as glucagon is unstable in solution and needs to be replaced more often, and more delivered insulin may be needed (62). In the past years, three outpatient clinical trials used a three-way comparison design between the dual-hormone (insulin and glucagon) CL, single-hormone CL and conventional insulin pump therapy (Table 1) (63 65). Only one of them showed a significant improvement of time the subjects spent within the target range with dual-hormone system when compared to single-hormone CL glucose control device (p = 0.032) (63). Another way that artificial pancreas devices can be classified is according to the type of computer algorithm that the digital controller uses. There are four main types of control algorithms being used in artificial pancreas systems. Model predictive control (MPC) algorithms predict glucose levels at a specific time point in the near future (some MPC-based systems can learn and adapt to the user s routine and make use of the clinician s input). Proportional integral derivative (PID) algorithms respond to measured glucose levels (66). Fuzzy logic algorithms calculate insulin doses based on how a clinical expert would make real-time adjustments based on CGM data (67). Lastly, bio-inspired algorithms are based on a mathematical model of how pancreatic beta cells produce insulin in response to changes in blood glucose levels. In addition to using CGM data, some artificial pancreas devices measure other biometric/physiological fluctuations (eg. galvanic skin response), and these are known as multivariable or adaptive systems. Artificial pancreas systems showed a greater time within the target range in overnight studies than in the studies done over 24 h. PID algorithms were associated with a lower difference in the time within the target range compared with MPC and fuzzy logic algorithms, although the tests for subgroup differences were not significant. PID (68). This finding is consistent with a recent trial directly comparing MPC and PID (66). In the past seven years, the evaluation of artificial pancreas use has been gradually transferred from controlled inpatient conditions (69,70), to controlled outpatient environments, such as hotels and summer camps, and included preschoolers, children, adolescents and adults of different ages (Table 1). There are some reports on artificial pancreas use during pregnancy and on the population of adults with type 2 diabetes (71,72). The current data almost unanimously supports the use of artificial pancreas as a safe and efficacious therapeutic option, with clinically relevant improvement in the time spent within the target range. The recent meta-analysis showed a clinically significant improvement of more than 12% of the time spent within the target range with the use of artificial pancreas devices compared to glucose control without the computer algorithm, without concurrent increase in risk of hypoglycemia or blood glucose excursions (68). As a result of this positive outcome, the artificial pancreas system was studied in a pivotal trial of a sufficient size (73) and is becoming a part of unsupervised clinical care (74). A greater consistency in reporting of basic outcome measures may even facilitate this process (75). 31

32 Table 1. The overview of randomized controlled trials on outpatient use of artificial pancreas in type 1 diabetes for glycemic control by year and length of observation period. OVERNIGHT STUDIES Year First author (Reference) Participants (n) Age (mean in years) Length of observation period Intervention Study utcomes Outcome difference: intervention vs. control p value 2013 Philip (76) night fuzzy logic/single time in target range 1.4 h < 0.05 time in hypoglycemia 0% 0.02 mean glucose -14 mg/dl < Hovorka (77) weeks MPC/single time in target range 15% < time in hypoglycemia -7% < 0.01 mean glucose -14 mg/dl < Nimri (78) weeks fuzzy logic/single time in target range 21.8% time in hypoglycemia -40.2% mean glucose -15 mg/dl Nimri (79) nights fuzzy logic/single time in target range 1.3 h time in hypoglycemia min mean glucose 3.5 mg/dl Ly (80) nights PID/single time in target range 7% time in hypoglycemia / / mean glucose 1 mg/dl Thabit (81) weeks MPC/single time in target range 12% time in hypoglycemia -0.3% 0.28 mean glucose mg/dl Brown (82) nights PID/single time in target range 26.30% < time in hypoglycemia -0.99% NS mean glucose mg/dl <

33 OVERNIGHT STUDIES Year First author (Reference) Participants (n) Age (mean in years) Length of observation period Intervention Study utcomes Outcome difference: intervention vs. control p value 2015 Haidar (63) nights MPC/dual/single time in target range dual CL vs. SAP 33% < single CL vs. SAP 16% time in hypoglycemia dual CL vs. SAP -1.7% single CL vs. SAP 0% Kropff (83) weeks MPC/single time in target range 8.6% < time in hypoglycemia -1.0% decrease in HbA1C -0.3% Thabit (84) weeks MPC/single time in target range 8.9% < time in hypoglycemia 0.82% 0.18 mean glucose -9 mg/dl Haidar (64) nights MPC/dual/single time in target range dual CL vs. SAP 22% < single CL vs. SAP 15% < time in hypoglycemia dual CL vs. SAP -7% < single CL vs. SAP 6% Ly (85) nights PID/single time in target range 15.8% time in hypoglycemia 14.1% mean glucose

34 OVERNIGHT STUDIES Year First author (Reference) Participants (n) Age (mean in years) Length of observation period Intervention Study utcomes Outcome difference: intervention vs. control p value 2016 Sharifi (86) (adults) 15.2 (children) 4 nights PID/single vs. LGS time in target range 6.2% 0.13 time in hypoglycemia 1.1% < mean glucose 2.0 mg/dl Nimri (87) nights fuzzy logic/single time in target range 13.5% time in hypoglycemia -0.53% mean glucose -7.9 mg/dl DAY AND NIGHT OBSERVATIONAL PERIOD Year First author (Reference) Participants (n) Age (mean in years) Length ob observation period Intervention Study outcomes Outcome difference: intervention vs. control 2014 Leelarathna (88) days MPC/single time in target range 13% time in hypoglycemia 1.3% mean glucose mg/dl Kovatchev (89) h PID/single time in target range -4.6% > 0.1 time in hypoglycemia -0.55% > 0.1 mean glucose 9 mg/dl < Russell (62) (adults) 16 (children) 5 days MPC/dual time in target range 20.7% (adults) 11.4% (children) < < time in hypoglycemia -3.2% (adults) -1.8% (children) mean glucose -26 mg/dl (adults) -16 mg/dl (children) <

35 DAY AND NIGHT OBSERVATIONAL PERIOD Year First author (Reference) Participants (n) Age (mean in years) Length ob observation period Intervention Study outcomes Outcome difference: intervention vs. control 2015 Thabit (84) weeks MPC/single time in target range 11% <0.001 time in hypoglycemia 0.8% 0.02 mean glucose -11 mg/dl < Ly (90) days PID/single vs. LGS time in target range -3.2% time in hypoglycemia -0.3% mean glucose 10 mg/dl De Bock (91) 8 Unknown 5 days PID/single vs. LGS time in target range 6.4% 0.30 time in hypoglycemia 0.1% 0.84 mean glucose mg/dl Blauw (92) days PID/dual time in target range 16.2% time in hypoglycemia -1.1% mean glucose -7.2 mg/dl Russell (93) days MPC/dual time in target range 23% < time in hypoglycemia 1.6% < mean glucose mg/dl Tauschmann (94) weeks MPC/single time in target range 18.8% <0.001 time in hypoglycemia 0.4% 0.33 mean glucose mg/dl Del Favaro (95) h MPC/single time in target range -6.3% time in hypoglycemia -4.7% <0.001 mean glucose 12 mg/dl <

36 DAY AND NIGHT OBSERVATIONAL PERIOD Year First author (Reference) Participants (n) Age (mean in years) Length ob observation period Intervention Study outcomes Outcome difference: intervention vs. control 2016 Tauschmann (96) days MPC/single time in target range 19% <0.01 time in hypoglycemia 1.2% 0.87 mean glucose mg/dl El Khatib (97) days MPC/dual time in target range 16.5% < time in hypoglycemia -1.3% < mean glucose mg/dl < Haidar (65) h MPC/dual time in target range dual CL vs. SAP 3.2% 0.31 single CL vs. SAP 3.0% 0.41 time in hypoglycemia dual CL vs. SAP -4.0% single CL vs. SAP -3.4% De Boer (98) days PID/single time in target range 26% <0.001 time in hypoglycemia -0.5% NS mean glucose -38 mg/dl < Bally (99) weeks MPC/single time in target range 10.6% < time in hypoglycemia -2.4% < mean glucose -7.2 mg/dl Table 1 legend. LGS low glucose suspend, MPC - model predictive control, NS not significant, PID proportional integrative derivative, SAP sensor augmented pump Single system with only insulin delivery. Dual system with both insulin and glucagon delivery. 36

37 Artificial pancreas and physical activity The data on artificial pancreas use with physical activity is scarce, especially in pediatric and adolescent population ( ). In 2013, an in-hospital randomized, crossover design trial compared the overnight blood glucose control during artificial pancreas and SAP after both sedentary and exercise days (60 minutes treadmill walking with 65 70% maximum heart rate) (103). Compared to SAP, there was no increase in the frequency of nocturnal hypoglycemia following exercise with artificial pancreas glucose control and the overall events frequency was lower (one nocturnal hypoglycemia with and one without antecedent exercise, compared to eight with and six without, p = 0.04). In this trial, children were instructed to achieve only about 50 to 55% of the maximal effort (equivalent to 65 70% maximum heart rate). Such research design does not reflect the patterns of physical activity observed in children s everyday activities (104). Incorporating short high-intensity sprints into intermittent exercise is also associated with less hypoglycemia following exercise (24, ) and is recommended by ISPAD as a possible strategy to minimize the risk of hypoglycemia (109). 37

38 Purposes of the thesis The purpose of the present thesis was to assess the cardiorespiratory fitness capacity of cohort of young people with type 1 diabetes, to determine whether glucose control with artificial pancreas use is safe and efficacious during an unannounced (to the artificial pancreas algorithm) afternoon exercise and the following afternoon/night until 13:00 of the next day, and to access the exercise-related metabolic changes in the cohort of young people with type 1 diabetes. 38

39 The aim of the thesis The objectives of this Ph.D. thesis were to to assess the cardiorespiratory fitness capacity of cohort of young people with type 1 diabetes, to evaluate efficacy, measured as proportion of time in glucose target range between 3.9 and 10.0 mmol/l (70 and 180 mg/dl), to evaluate safety, measured as the proportion of time in hypoglycemia below 3.3 mmol/l (60 mg/dl), proportion of time below 3.9 mmol/l (70 mg/dl) and the hypoglycemia event rate of artificial pancreas glucose control when compared to SAP during the afternoon exercise and the afternoon/night after (till 13:00 next day) based on sensor glucose readings, and to assess the exercise-related metabolic changes in the cohort of young people with type 1 diabetes during the afternoon exercise and on the morning of the following day. The thesis hypotheses were: 1. The cardiorespiratory capacity of our cohort of children and adolescents with type 1 diabetes is similar to their healthy peers. 2. The artificial pancreas will increase the proportion of time inside the target range between 3.9 and 10.0 mmol/l when compared to SAP. 3. The artificial pancreas will reduce the proportion of time inside the hypoglycemia below 3.3 mmol/l (60 mg/dl), below 3.9 mmol/l (70 mg/dl) and the hypoglycemia event rate when compared to SAP. 4. Exercise-related metabolic respons will differ when challenged by exercise protocol of moderate physical activity with high-intensity sprints vs. without high-intensity sprints. Methods Present clinical trial adopted an open-label, randomized, in-hospital, two-arm crossover design contrasting glycemic control with artificial pancreas to glycemic control without the computer algorithm decision-making. Each clinical trial intervention session lasted for 24 h, with a one-week washout period between the two interventions. Participants We recruited and enrolled the clinical trial participants between January and May 2016 through Slovenian National Diabetes Registry (7) at University Children s Hospital in Ljubljana. 39

40 Inclusion criteria All eligible participants fulfilled the following inclusing criteria: 1. The participant had a documented T1D, as defined by the American Diabetes Association and World Health Organization for at least 1 year prior to enrolment. 2. The participant had been using an insulin pump for at least 3 months. 3. The participant was between 10 and 17 years of age (inclusive) at the time of enrolment. 4. The participant had an HbA1c value equal or below 9% at the time of enrolment. 5. The participant was willing and able to follow all clinical trial instructions (the child and their parent were evaluated as a unit). 6. The participant was available for the entire clinical trial duration and follow-up visits. 7. The participant was willing to perform daily SMBG and the required sensor calibrations. 8. If the participant had celiac/hashimoto disease, the disease had to be adequately under control as determined by the investigator. 9. The participant had a BMI within the normal range for age and gender (± 2 SD). Exclusion criteria 1. Concomitant diseases that influence metabolic control (e.g. anemia, significantly impaired hepatic function, renal failure, history of adrenal insufficiency) or other medical conditions, which in the investigator s opinion, could have compromised the patient s safety (with the exception of adequately treated autoimmune thyroid and celiac disease). 2. Significant co-morbidity that, by the investigators opinion, would preclude participation in the clinical trial (e.g. current treatment for cancer, mental disorder). 3. The participant was taking or has taken oral or parenteral glucocorticoids within 1 month prior to screening, or is planned to take oral or parenteral glucocorticoids during the clinical trial. Exceptions: short term oral glucocorticoids up to 7 days, inhaled steroids. 4. The participant was taking antidiabetic agents or other medications, which could be a contraindication to the participation in the study as judged by the investigator. 5. The participant was taking part in another clinical trial of a medical device or drug that could affect glucose measurements or glucose management, or a recipient of any investigational medical product within 1 month prior to the screening (Visit 1). 6. A female participant of child-bearing potential who was pregnant, breastfeeding, or was planning to become pregnant during the clinical trial. 7. The participant had known hypoglycemic unawareness or recurrent severe hypoglycemic events with seizure and/or coma (more than two episodes) within 6 months prior to screening. 8. The participant had a history of one or more episodes of DKA requiring hospitalization within a month prior to screening. 9. The participant had current or recent history of alcohol or drug abuse. 10. The participant had visual impairment or hearing loss, which in the investigator s opinion, could have compromised the patient s ability to perform the clinical trial procedures safely (the child and their parent were evaluated as a unit). 40

41 Procedures All participants came to the clinic on six occasions: the first two for education, screening, randomization, and baseline assessment, followed by four 24-hour in-patient sessions, from noon (12:00) untill 13:00 of the following day. The screening visit (visit 1) included an informed consent acquisition, detailed physical examination, and a confirmation of the inclusion/exclusion criteria. The participants were trained in the use of the glucose sensor before entering the run-in period. At the baseline visit (visit 2), all participants performed a lung function test, resting ECG and a cycle ergometer exercise test to determine their maximum oxygen consumption rate (VO 2max ). We downloaded the run-in period (at least 5 days) glucose sensor data to derive the initial personalized artificial pancreas system settings (see below for details) and to optimize pump settings that remained the same for all study visits. The participants were instructed to refrain from intensive physical activity for 48 hours prior to the four in-patient visits. For the in-patient visits (visits 3 6), the participants were randomized to complete the first two sessions in the artificial pancreas group first, followed by two visits in the control group, or vice versa. The insulin dose during the in-hospital stay was decided by a trained nurse educator in consultation with the participant and, if needed, with the on-call diabetologist. On the basis of the insulin pump data, we excluded the participants with a hypoglycemic event (glucose level below 2.8 mmol/l) on the day before the intervention, as they could be at an ncreased risk of hypoglycemia during the exercise. The participants were instructed to insert the sensor on the day before their hospital visit. Upon admission, a nurse educator checked the sensor and a backup sensor was inserted if needed. Whole observational period was defined from 15:00 on an exercise day untill 13:00 on the following day. Overnight period was from 23:00 untill 7:00 on the folowing day. Exercise started in the time period from 16:30 to 19:30, lasted for 40 minutes and additional 4 hours (2 hours with frequent SMBG) were included into exercise period data analysis. To detect hypoglycemia on the day following exercise we prolonged our observational period until 13:00 h the next day (31). The schedule for all exercise days was similar: - 12:00 hospital admission - 13:00 lunch - 13:30 to 16:00 rest - 15:00 to 16:00 afternoon snack - 16:30 to 19:30 cycle ergometer exercise protocol - Dinner one hour after the exercise - 22:00 to 7:30 next day: bedtime - 8:00 breakfast - 13:00 lunch and the end of observation period 41

42 During the in-patient stay, the sensor calibration was regularly scheduled at three different times: at the initiation of the artificial pancreas insulin delivery system (around 15:00 h on day 1), before dinner (from 20:00 to 21:00 h) and in the morning of day two (around 7:00 h). If the glucose level was out of range, or if there was a discrepancy between the blood and sensor glucose, the calibration was postponed until the glucose level was stable. Exercise protocol Each hospitalization participant performed a 40-min of afternoon physical activity on a cycle ergometer. Two different exercise protocols were carried out once in each glucose control group: a moderate intensity (55% VO 2max ) physical activity or a combination of a moderate physical activity with incorporated high-intensity (80% VO 2max ) sprints (55/80% VO 2max ). In both protocols, the participants were instructed to pedal at a steady rate of rev/min for 40 min, at 55% VO 2max load (the load was achieved after about five minutes and was based on VO 2max adjusted in real time as needed to maintain 55% VO 2max level). In the 55/80% VO 2max protocol, high-intensity sprints with a duration of 20 s were incorporated, with intervals of 6 10 min activity at 55% VO 2max between the sprints, for a total of 40 min. Throughout the exercise, a continuous ECG was recorded, and the inhaled O 2 and exhaled CO 2 were measured. Capillary blood glucose was checked at the beginning of each exercise session, every 15 min during the exercise and every 30 min for 2 h following the exercise, blood lactate was measured at the beginning and at the end of exercise protocol, and fasting cortisol level was measured in the moring of the following day to determin possible overreaching with the exercise protocol ( ). Standardized meals and insulin delivery During the hospitalization, all participants received standardized meals containing approximately 1 gram of carbohydrates per kg of body mass for the main meals (lunch, dinner, and breakfast) and about half of this amount for the snack. Participants got the same meals for all 4 study visits. The amount of carbohydrates, proteins and fat was predetermined. Carbohydrates intake and calories were based on the patient s weigh and were designed by a dietitian to contain 50% carbohydrate, 20% protein, and 30% fat, of which no more than 10% saturated fat, following the American Diabetes Association recommendations (23). In both study arms, all meals were covered with the manual insulin boluses according to the individual s carbohydrate-to-insulin ratio that remained the same for all study visits. For the control group, the insulin pump was disconnected during the exercise and the basal insulin dose was reduced by 20% for 4 h following the exercise session. 42

43 During the artificial pancreas insulin delivery, the use of the pump was uninterrupted; the device was applied from 15:00 h on the day of the exercise to 13:00 h of the following day, and the exercise was not announced to the artificial pancreas control algorithm. Devices and Assays During the clinical trial period, all participants used the identical insulin pump (Paradigm Veo, Medtronic Diabetes, Northridge, CA), a real-time CGM (Enlite sensor with MiniLink REAL time transmitter, Medtronic Diabetes) and a glucose meter (Contour Link meter, Ascensia HealthCare, Basel, Switzerland). The»suspend on low«and»suspend before low«functions were disabled for all the participants. The exercise protocol was performed on a cycle ergometer (Power Cube LF8.5G with Schiller software; Ganshorn, Niederlauer, Germany). The HbA1c level was determined by an immunochemical method using the Siemens DCA Vantage Analyser and lactate and cortisol with Advia Centaur (Siemens Healthcare, USA). Fuzzy logic artificial pancreas DreaMed Glucositter In the present clinical trial, we evaluated the Dreamed Glucositter (DreaMed Diabetes, Petah Tikva, Israel) artificial pancreas device fuzzy-logic based control algorithm (software version ), which operated on a commercial laptop/tablet computer (ThinkPad T450s; Lenovo, Beijing, China). It used a modified vendor-supplied communication module application programming interface (API) in order to retrieve the glucose/insulin data from the MiniMed Paradigm Veo with a physical connection to a communication dongle (provided by the manufacturer of the insulin pump) and it set the insulin treatment accordingly. The artificial pancreas software was implemented using the Matlab platform (MathWorks, Natick, MA, USA) and includes the following features: (i) a communication module, (ii) user interface, (iii) a control algorithm, and (iv) an alert module (Figure 3). The DreaMed Glucositter supports communication with the Paradigm Veo (Medtronic Co.) The Paradigm Veo System includes an insulin pump, a CGM (subcutaneous wired sensor, named SOF sensor or ENLITE sensor), a Minilink transmitter, a ComLink (wireless, low-powered, radio frequency (RF) communication suitable for European and US frequencies) and a RS232 to USB converter cable (Figure 4). The main logic behind the system was the use of the reasoning of the diabetes caregivers and traditional treatment principles. It uses a fuzzy logic technology in order to mimic this reasoning. Therefore, the system uses a basal/bolus insulin dosing regimen, the individual patient s treatment management (which set the patient s sensitivity to insulin, similar to the way the physicians 43

44 sets it for the patients on the insulin pump treatment). In addition, as a physician changes the insulin sensitivity from time to time, the system uses a learning algorithm that can change the patient s treatment profile. Fuzzy logic is the science of reasoning, thinking, and inference, which recognizes that not everything is true or false in the real world. According to fuzzy logic, the correctness of any statement becomes a matter of degree. The main elements of the fuzzy logic controller are fuzzy sets of multiple inputs and single or multiple outputs, fuzzy rules structured according to the form of IF (input)-then (output), and methods of fuzzification and defuzzification to evaluate the fuzzy-rule output based on the input. Figure 3. A general description of the MD-Logic artificial pancreas. Figure 4. A sensor-augmented insulin pump components. 44

45 The CGM steadily measures the interstitial glucose levels and these values are sent by the Minilink via wireless, low-powered, radio frequency (RF) to the Paradigm Veo every 5 minutes. The blood glucose (BG) (Contour Link) meter is used to calibrate the readings measured by the CGM. The Paradigm Veo displays and stores the glucose measurements sent by the Minilink. No modifications have been made to either the sensor or the insulin pump. The insulin pump communicates to the host computer via the Comlink (wireless, low-powered, radio frequency (RF) communication). The artificial pancreas requires a patient-specific log file for its operation. This log file includes the treatment settings for each individual that are downloaded from the SAP (based on run-in period data an individual s sensitivity factor, carbohydrate factor and basal insulin settings). Once this pre-made log file exists inside the artificial pancreas device (a dedicated laptop in this case) for each individual, the physician can launch the application, check and approve the settings and insert the pump serial number. From there, the system automatically connects to the pump and sensor and controls the insulin dosing. The controller (Figure 5) applies a combination of the two control strategies: control to range and control to target. The aim of the control to range strategy is to bring the patient s glucose levels to within the desired range; the aim of the control to target strategy is to bring the patient s glucose to a specific target level. The control to range fuzzy logic controller uses treatment rules that were phrased in collaboration with the medical staff. The declared goal of the rules was to keep the glucose levels stable within target range. The rules use four inputs that are calculated from CGM readings: past and future glucose trend as well as current and future glucose levels. Each rule has two outputs: 1) change in basal rate and 2) portion of insulin bolus (in percentage from the patient s basal plan and the calculated bolus). The control-to-target module aims to bring the patient s glucose to a specific target level. To reach the final dosing recommendation, the control to target modul takes into consideration the 1) recommendation of the Control to range modul (in percentage), 2) the predefined glucose target level, 3) insulin dosing regimen history, and 4) safety constraints related to the insulin pharmacodynamics (69). Both the patient s treatment profile and the performance of the controller are adjustable, making it easier for the system to deal with the inter- and intra-patient variability. 45

46 Figure 5. A shematic description of the control algorithm. Clinical data of artificial pancreas device DreaMed Glucositter The scope of the artificial pancreas device used in the present clinical trial has previously been evaluated in several clinical trials. In the pilot study conducted in 2010 seven adults underwent 14 full artificial pancreas control sessions of 8 h (with fasting and meal challenge conditions) and two additional 24-hour visits. During the 24-hour artificial pancreas control, they spent 73% of time within the range ( mmol/l), 27% above 10 mmol/l, none below 3.9 mmol/l and without symptomatic hypoglycemic events (69). In a multicenter in-patient study, 12 eligible participants (with three additional participants for the pilot phase of the study) were enrolled from the three participating centers (age 23.8±15.6 years, HbA1c 8.1±0.8%). The proportion of time spent in the target range of mmol/l was significantly increased with the artificial pancreas use by 20% (p = 0.002) compared to CSII and with a significantly less glucose variability as measured by HBGI (p = 0.02) (114). A multinational, randomized, crossover clinical trial assessed the short-term safety and efficacy of the device for the control of nocturnal glucose levels in children with type 1 diabetes at a diabetes camp. 56 young patients were randomized in two consecutive overnight sessions with 46

47 artificial pancreas and SAP glucose control in random order and the data from 54 participants was included in the ITT analysis (age 13.8±1.8 years, HbA1c 8.0±0.7%). The artificial pancreas use reduced the number of sensor glucose values below 3.5 mmol/l (7 vs. 22, p = 0.003), shortened the time spent with glucose levels below 3.3 mmol/l (p = 0.02), and increased the time spent within the target range mmol/l for 1.4 h (p < 0.05) of the night (76). In a multinational, multicenter, randomized clinical trial 75 participants (aged years; average HbA1c 7.8 ± 0.7%, 61.8 ± 7.2 mmol/mol) were randomly assigned to participate in two overnight crossover periods, each including four consecutive nights with both artificial pancreas and SAP glucose control. The study results showed that during the nights when the artificial pancreas was used, the time spent in hypoglycemia was significantly reduced (p = 0.004) compared to the nights when the SAP therapy was used instead. The percentage of individual nights in which mean overnight glucose level was within mmol/l was significantly increased by 25% (p = 0.008) (87). The longest outpatient clinical trial with the present artificial pancreas device evaluated the effect on overnight glycemic control at the patients home for a period of 6 weeks (78). Twenty-four patients (aged years, average HbA1c 7.5 ± 0.8%, 58.1 ± 8.4 mmol/mol) were randomly assigned to participate in two overnight crossover periods each for 6 weeks of consecutive nights: one under artificial pancreas glucose control and the second under SAP therapy at patients home in real-life conditions. The primary endpoint was the time spent with sensor glucose levels below 3.9 mmol/l overnight. The study results of intention-to-treat analysis showed that the use of artificial pancreas during the night significantly reduced the time spent in hypoglycemia for 1.9% (p = 0.02) and increased the proportion of time spent in the target range of mmol/l for 13.5% (p = 0.003) compared to the nights when the sensor-augmented pump therapy was used instead. The total overnight insulin doses were lower for the artificial pancreas nights compared to the sensor-augmented pump nights (p = 0.04). No serious adverse events were reported in any of the conducted studies. None of the previous studies with DreaMed Glucositter incorporated specific guidelines relating physical activities. Randomization and masking The participants were randomly assigned, using 1:1 ratio, to perform a physical activity for two consecutive days using the artificial pancreas insulin delivery, followed by the physical activity for two days using an insulin pump with a glucose sensor and without the computer algorithm (control group), or vice versa. Following the run-in period, the order was randomly determined using an automated web-based program with randomly permuted blocks of four. The participants and the investigators analyzing the study data were not blinded to treatment. 47

48 Statistical analyses The statistical analysis was based on the modified intention-to-treat (ITT) population, defined as all randomly assigned participants with more than 67% sensor measurements. The comparisons between the artificial pancreas and control were performed using the paired nonparametric Wilcoxon signed rank test. The power of the nonparametric tests for the primary endpoint was estimated on the basis of the results of power simulations (MATLAB 2013b, MathWorks), which were performed using sensor data from 102 participants with the eligibility profile under standard treatment at home. On the basis of previous in- (115) and outpatient (76,79) studies, we calculated that the enrolment of 20 participants in total would provide a power of 90% for detecting a 30% reduction in the time below 3.3 mmol (60 mg/dl) at a 0.05 two-sided significant level, assuming 30% dropout. Safety monitoring A hypoglycemic event was defined as a blood glucose level below 3.3 mmol/l based on the sensor glucose readings, with a minimum duration of 20 minutes. All sensor glucose values of less than 3.3 mmol/l were recorded by the study device and included in the statistical analysis. For this clinical trial, only hypoglycemic events meeting the specific criteria listed bellow were reported as an adverse event. All sensor glucose-detected hypoglycemic events were additionally confirmed with the SMBG. When the glucose values fell below 3.3 mmol/l if symptomatic, and when the glucose values fell below 2.8 mmol/l, regardless of the symptoms, 15 grams rescue carbohydrates were administered as per standard in-hospital procedures and recorded as an adverse event, so not all sensor glucose-detected hypoglycemic events were treated with rescue carbohydrates. Severe hypoglycemia was considered a serious adverse event and was defined as glucose under 2.8 mmol/l, accompanied by seizures or loss of consciousness, as per ISPAD guidelines, or if it required intravenous glucose and/or intramuscular glucagon administration. Hyperglycemia or diabetic ketoacidosis were considered a serious adverse event only if glucose level was above 13.9 mmol/l and was associated with low serum bicarbonate (<15 mmol/l) or low ph (<7.3) and either ketonaemia (β- hydroxybutyrate level above 3 mmol/l) or ketonuria requiring intravenous treatment. Other hyperglycemic events were not reported as adverse events; however, they were recorded by the study device and included in the final analysis. Clinical trial Management and Regulatory Approvals The clinical trial was conducted in compliance with the protocol, the Declaration of Helsinki, and applicable regulatory and Good Clinical Practice requirements, with the approval of the Slovenian National Ethics Committee No. KME 135/06/15 and Slovenian regulatory agency (Agency for Medicinal Products and Medical Devices, Slovenia No. CIV-Sl ). All participants and their parents provided a written informed consent prior to the trial initiation. The clinical trial protocol is listed on clinicaltrials.gov under the registration number NCT

49 Thesis Results After an extensive preparatory work based on the data analysis from the Sloverne national childhood-onset type 1 diabetes registry, which was published separately (7), eligibility criteria were selected. Between 20 January and 16 June 2016, 25 children and adolescents with type 1 diabetes were invited to participate, through the Slovenian National Diabetes Registry, and 20 (nine female) were included and randomized. They all completed the clinical trial and provided data for analysis (study flow diagram is presented in Figure 6), published in the reference (116). Enrollment Assessed for Eligibility (n=25) Excluded (n=5) Not meeting inclusion criteria (n=1) Declined to participate (n=4) Allocation Randomized (n=20) Closed-loop (n=10) Completed intervention (n=10) Crossover Open-loop (n=10) Completed intervention (n=10) Closed-loop (n=10) Completed intervention (n=10) Open-loop (n=10) Completed intervention (n=10) Analysis Analysed (n=20) Excluded from analysis (n=0) Figure 6. Study flowchart. 49

50 Characteristics of participants Baseline characteristics for whole study population and separate for each gender are presented in Table 2. The mean age of all participants was 14.2 ± 2.0 years, duration of diabetes 8.3 ± 3.2 years, HbA1c 7.7 ± 0.6% (60.0 ± 6.6 mmol/mol), duration of pump therapy 7.4 ± 3.2 years and the total daily insulin dose 0.8 ± 0.2 U/kg. Their BMI (21.5 ± 4.3 kg/m2) was within normal range. They were of average glycemic control with HbA1c at inclusion 7.7±0.6% (60±6.6 mmol/mol). Girls were on average slightly older and with longer duration of type 1 diabetes and insulin pump treatment than boys. Their daily insulin need was the same. Table 2. Participants characteristics. Characteristic All (n=20) Male (n=11) Female (n=9) Age (years) 14.2± ± ±2.0 Duration of diabetes (years) 8.3± ± ±3.8 Duration using pump (years) 7.4± ± ±3.8 BMI (kg/m 2 ) 21.5± ± ±4.2 BMI SDS (percentile) 63.6± ± ±21.8 HbA 1c (%) 7.7± ± ±0.7 HbA 1c (mmol/mol) 60± ± ±7.7 Total daily insulin (U/kg) 0.8± ± ±0.2 Cardiorespiratory fitness capacity At the baseline visit, all participants performed a lung function test and a cardiovascular fitness assessment with a maximal load test (Table 3). They were found to be of average cardiorespiratory fitness: VO 2max 43.3 ± 9.3 ml/kg -1 min -1 (36.1 ± 4.0 ml/kg -1 min -1 for girls and 49.2 ± 8.1 ml/ kg -1 min -1 for boys) and maximal heart rate ± 10.2 beats/min (182.9 ± 11.9 beats/min for girls and ± 7.9 beats/min for boys). At the end of the maximal load test their lactate level was 7.8±2.2 mmol/l (8.8±2.4 mmol/l for boys and 6.9±1.7 mmol/l for girls). Table 3. Lung function test results and cardiorespiratory fitness capacity. All (n=20) Male (n=11) Female (n=9) FEV1 (L) 3.3± ± ±2.0 Maximal load (Watts) 169.8± ± ±30.9 VO 2max (L) 2.4± ± ±0.4 VO 2max (ml/kg -1 min -1 ) 43.3± ± ±4.0 HR end (beats/min) 186.6± ± ±11.9 Lactate start (mmol/l) 1.2± ± ±0.4 Lactate end (mmol) 7.8± ± ±1.7 FEV1 Forced Expiratory Volume in 1 second, HR Maximal Heart Rate at Maximal Load Test, VO 2max Maximal Oxygen Capacity, Lactate start and end Lactate Value at the Beginning and at the End of Maximal Load Test. 50

51 Glucose control The data representing glucose control is shown in Tables 4 9 and Figures 7 8. For the total duration of the clinical trial, we obtained 96% of sensor data during the artificial pancreas delivery and 97% SAP (control) delivery. The data from all participants was included in the analysis. The amount of carbohydrates consumed during the study period was 1.14 ± 0.34 g/kg for main meals and 0.55 ± 0.41 g/kg for snacks. Figure 7. Median (IQR) sensor glucose during artificial pancreas (blue) and SAP (red) insulin delivery, from the start of the observational period (17:00 h) until (13:00) following day, with standard glucose outcome measures (75) and a common glucose target value of 6.1 mmol/l. SAP Sensor Augmented Pump (without computer algorithm). Time in target range The artificial pancreas delivery of insulin increased the proportion of time spent within the glucose range of mmol/l when compared to SAP delivery: 84.1% ( ) vs. 68.7% ( ), p = (Table 4). This was also true when calculated for the nightime alone: 92.8% ( ) for the artificial pancreas delivery and 73.3% ( ) for SAP, p = A similar difference was observed for tighter glucose range mmol/l, where the artificial pancreas use increased the time in range for 8.2%, p = and for 1.5% at night p = The median and mean glucose levels for the whole observational period were lower during the artificial pancreas delivery of insulin than during SAP (p = and p = , respectively). There was no difference in glucose variability between the treatment modalities as measured by the mean SD. 51

52 Table 4. Comparison of glucose control during artificial pancreas and SAP insulin delivery for outcome measures of near normoglycemia. Variable Artificial pancreas SAP (control) p value Proportion of time spent with glucose within mmol/l (%) All 84.1 ( ) 68.7 ( ) Night 92.8 ( ) 73.3 ( ) % 80.9 ( ) 68.1 ( ) /80% 75.3 ( ) 68.4 ( ) Proportion of time spent with glucose within mmol/l (%) All 26.5 ( ) 18.3 ( ) Night 17.0 ( ) 15.5 ( ) % 22.1 ( ) 15.4 ( ) /80% 23.0 ( ) 14.8 ( ) Median glucose (nmol/l) All 7.5 ( ) 8.6 ( ) Night 7.8 ( ) 8.1 ( ) % 7.8 ( ) 8.7 ( ) /80% 7.7 ( ) 8.5 ( ) Mean glucose (mmol/l) All 7.9 ( ) 8.8 ( ) Night 7.8 ( ) 8.6 ( ) % 8.0 ( ) 8.7 ( ) /80% 7.9 ( ) 9.3 ( ) Mean SD of glucose concentration 41.1 ( ) 44.6 ( ) Fasting glucose (mmol/l), all 6.8 ( ) 7.3 ( ) Data is shown as the median (IQR). Nonparametric data analysis of glucose control (paired nonparametric Wilcoxon signed rank test) and outcome data are presented as median (IQR) although the variables were presented as mean (glucose, glucose concentration SD) All: whole study period from 15:00 h on day of exercise until 13:00 h on the following day; Night: period from 22:00 h on day of exercise until 07:00 h on the following day 55%, exercise protocol with moderate physical activity; 55/80%, exercise protocol with moderate physical activity and high-intensity sprints; SAP Sensor Augmented Pump (without computer algorithm). Time in hypoglycemia The median (interquartile range) proportion of time spent in hypoglycemia below 3.3 mmol/l for the whole observational period based on sensor glucose readings, was 0.00% for both treatment modalities ( % for artificial pancreas and % for SAP, p = ) (Table 5). During the study period, six hypoglycemic events were recorded in the artificial pancreas group 52

53 and 12 in SAP (p = ); the participants received rescue carbohydrates on seven occasions in the artificial pancreas group (total of 105 g) and on eight occasions in the SAP (total of 120 g) (Table 10). There were no significant differences between the two groups for other variables of hypoglycemia (i.e. proportion of time spent with glucose below 3.9 mmol/l, AUC below 3.3 mmol/l and 3.9 mmol/l, and low blood glucose index (LBGI)). Table 5. Comparison of glucose control during artificial pancreas and SAP (control) for the outcome measures of hypoglycemia. Variable Artificial pancreas SAP (Control) p value Time spent at low glucose Proportion of time spent with glucose below 3.3 mmol/l (%) All 0.0 ( ) 0.0 ( ) Night 0.0 ( ) 0.0 ( ) % 0.0 ( ) 0.0 ( ) /80% 0.0 ( ) 0.0 ( ) Proportion of time spent with glucose below 3.9 mmol/l (%) All 0.8 ( ) 1.1 ( ) Night 0.0 ( ) 0.0 ( ) % 0.0 ( ) 0.2 ( ) /80% 1.1 ( ) 0.0 ( ) Glucose AUC below 3.3 mmol/l (mmol/l min) All 0.0 ( ) 0.0 ( ) Night 0.0 ( ) 0.0 ( ) % 0.0 ( ) 0.0 ( ) /80% 0.0 ( ) 0.0 ( ) Glucose AUC below 3.9 mmol/l (mmol/l min) LBGI All 9.2 ( ) 8.6 ( ) Night 0.0 ( ) 0.0 ( ) % 0.0 ( ) 0.1 ( ) /80% 2.9 ( ) 0.0 ( ) All 0.4 ( ) 0.4 ( ) Night 0.3 ( ) 0.0 ( ) Data is shown as the median (IQR). All: whole study period from 15:00 h on day of exercise until 13:00 h on the following day; Night: period from 22:00 h on day of exercise until 07:00 h on the following day. 55%, exercise protocol with moderate physical activity; 55/80%, exercise protocol with moderate physical activity and high-intensity sprints; LBGI low blood glucose index, SAP Sensor Augmented Pump (without computer algorithm). 53

54 Time in hyperglycemia The use of artificial pancreas delivery of insulin significantly reduced the proportion of the time spent in hyperglycemia: by 13.5% for the time above 10.0 mmol/l (p = vs. SAP) and by 0.9% above 13.9 mmol/l (p = vs. SAP). The High blood glucose index (HBGI) was lower when the artificial pancreas delivery was used for the whole observational period (3.6 vs. 6.2, p = ) and for the nighttime alone (3.2 vs. 5.5, p = ) (Table 6). Table 6. Comparison of glucose control during artificial pancreas use and SAP (control) insulin delivery for the outcome measures of hyperglycemia. Variable Artificial pancreas SAP (control) p value Proportion of time spent with glucose above 13.9 mmol/l (%) All 1.3 ( ) 2.2 ( ) Night 0.0 ( ) 0.0 ( ) % 0.0 ( ) 0.0 ( ) /80% 2.3 ( ) 1.1 ( ) Proportion of time spent with glucose above 10 mmol/l (%) All 14.6 ( ) 28.1 ( ) Night 7.2 ( ) 22.7 ( ) % 17.1 ( ) 25.2 ( ) /80% 20.8 ( ) 29.6 ( ) Glucose AUC above 13.9 mmol/l (mmol/l min) All 17.3 ( ) 46.1 ( ) Night 0.0 ( ) 0.0 ( ) % 0.0 ( ) 0.0 ( ) /80% 11.4 ( ) 1.3 ( ) AUC above 10 mmol/l (mmol/l min) All ( ) ( ) Night ( ) 11,226.9 ( ,212.0) % ( ) ( ) /80% ( ) ( ) HBGI All 3.6 ( ) 6.2 ( ) Night 3.2 ( ) 5.5 ( ) Data is shown as the median (IQR). All: whole study period from 15:00 h on day of exercise until 13:00 h on the following day; Night: period from 22:00 h on day of exercise until 07:00 h on the following day. 55%, exercise protocol with moderate physical activity; 55/80%, exercise protocol with moderate physical activity and highintensity sprints; HBGI high blood glucose index, SAP Sensor Augmented Pump (without computer algorithm). 54

55 Glucose control during exercise period The SMBG data on glucose control during exercise and early recovery time is shown in Table 7A and the data on sensor glucose analysis in Table 7B. The amount of active insulin at the start of the exercise and the blood glucose values at the start and at the end of the exercise were similar for both treatment modalities (6.2±4.7 for artificial pancreas and 5.6±3.0, p = ), the same was true for the change in blood glucose levels from the start to the end of exercise: 2.6 ( 4.6 to 1.6) mmol/l for artificial pancreas insulin delivery and 2.2 ( 3.5 to 0.9) mmol/l for SAP (p = ). During exercise, there was one hypoglycemic event in the artificial pancreas group compared to four in the open-loop group (p = ). The time within the target range mmol/l was increased with artificial pancreas use by 21.4% (p = ), there was no difference in the time spent within hypoglycemia below 3.9 mmol/l (0.0% for both treatment modalities). The proportion of time in hyperglycemia above 10 mmol/l (p = ) and 13.9 mmol/l (p = ) was significantly reduced (Table 7B). 55

56 Table 7A. Analysis of blood glucose values during the exercise period and for 2 hours after. Blood glucose (mmol/l) p value Period Artificial pancreas SAP (Control) Physical activity start 8.3 ( ) 9.2 ( ) Start + 15 min 8.2 ( ) 8.9 ( ) Start + 30 min 7.0 ( ) 7.5 ( ) Physical activity end 5.8 ( ) 7.2 ( ) Δ Start to end 2.6 ( 4.6 to 1.6) 2.2 ( 3.5 to 0.9) End + 30 min 6.3 ( ) 8.4 ( ) End + 60 min 7.0 ( ) 9.5 ( ) End + 90 min 8.0 ( ) 10.4 ( ) End min 8.5 ( ) 11.0 ( ) Table 7B. Analysis of sensor glucose during the exercise period and four hours after. Variable Artificial pancreas SAP (control) p value Proportions of time (%) Below 3.3 mmol/l 0.0 ( ) 0.0 ( ) Below 3.9 mmol/l 0.0 ( ) 0.0 ( ) Within mmol/l 83.9 ( ) 62.5 ( ) Above 10 mmol/l 10.4 ( ) 33.2 ( ) Above 13.9 mmol/l 0.0 ( ) 0.0 ( ) Median glucose (mmol/l) 7.2 ( ) 8.6 ( ) Mean glucose (mmol/l) 7.4 ( ) 8.8 ( ) No. of hypoglycemia events Active insulin at start of exercise (U) 6.2± ± Data is shown as the median (IQR) and mean with SD. SAP Sensor augmented pump (without computer algorithm). Day two glucose control The data on glucose control for the day following physical activity is presented in Table 8. The difference between the study arms in the proportion of time spent with glucose within the range mmol/l favoring artificial pancreas delivery of insulin did not reach statistical significance (72.8% in artificial pancreas vs. 65.5% SAP, p = ). The proportion of time spent in hypoglycemia below 3.3 mmol/l (0.0% for both treatment modalities) and below 3.9 mmol/l (0.0% for both treatment modalities) was low, there was no difference between the two treatment modalities in the median and mean glucose values. 56

57 Table 8. Glucose control on the day following physical activity (7h 3h). Variable Artificial pancreas SAP p value Proportions of time (%) Below 3.3 mmol/l 0.0 ( ) 0.0 ( ) Below 3.9 mmol/l 0.0 ( ) 0.0 ( ) Within mmol/l 72.8 ( ) 65.5 ( ) Above 10 mmol/l 26.5 ( ) 28.9 ( ) Above 13.9 mmol/l 0.0 ( ) 0.0 ( ) Median glucose (mmol/l) 8.7 ( ) 8.6 ( ) Mean glucose (mmol/l) 8.5 ( ) 8.5 ( ) Data is shown as the median (IQR). SAP Sensor augmented pump (without computer algorithm). Glucose control during different exercise protocols In the subanalysis of glucose control with the artificial pancreas comparing the effects of different physical activity protocols on the time spent at various glucose concentrations, we observed a significant improvement during the 55/80% VO 2max protocol in the proportion of time spent within the target range mmol/l for 15.4% (p = ), but not during the 55% VO 2max protocol. There was no difference between the two treatment modalities in measures of hypoglycemia for any of the exercise protocols. Both the mean and median glucose levels were lower with the artificial pancreas use during the 55/80% VO 2max protocol and the time spent above 10 mmol/l (for 8.8%, p = ) was reduced (Table 4), but not during the 55% VO 2max protocol. Figure 8A. Median (IQR) sensor glucose during artificial pancreas (blue) and SAP (red) insulin delivery, for the whole observational period from 15:00 h until 13:00 h on the folowing day, for 55% VO 2max moderate physical activity protocol with consensus glucose outcome measures (75) and mean glucose target value 6.1 mmol/l. SAP Sensor augmented pump (without computer algorithm). 57

58 Figure 8B. Median (IQR) sensor glucose during artificial pancreas (blue) and SAP (red) insulin delivery, for the whole observational period from 15:00 h until 13:00 h on the folowing day, for 55/80% VO 2max protocol of moderate physical activity with high-intensity sprints with consensus glucose outcome measures (75) and mean glucose target value 6.1 mmol/l. SAP Sensor augmented pump (without computer algorithm). Insulin delivery During the artificial pancreas glucose control the device delivered a significantly lower total amount of insulin (112.6 IU vs IU with SAP, p = ), mostly on account of less basal insulin delivered (80.3 U vs U with SAP, p = ), with no difference in amount of bolus insulin (36.4 IU via artificial pancreas and 40.3 IU via SAP, p = ) (Table 9). Table 9. Comparison of insulin delivery: artificial pancreas and SAP over whole period. Insulin delivered Artificial pancreas SAP (control) p value Total daily insulin (U) All ( ) ( ) Night 96.2 ( ) ( ) Bolus insulin (U) 36.4 ( ) 40.3 ( ) Basal insulin (U) 80.3 ( ) ( ) Data are medians (IQR). All: whole study period from 15:00 h on day of exercise until 13:00 h on the following day; Night: period from 22:00 h on day of exercise until 07:00 h on the following day. SAP Sensor augmented pump (without computer algorithm). 58